glam/data/entity_annotation/entity_annotation_rules_v1.6.0_unified.yaml

# =============================================================================
# ⚠️ DEPRECATED - USE ch_annotator-v1_7_0.yaml INSTEAD ⚠️
# =============================================================================
# This file has been renamed to: ch_annotator-v1_7_0.yaml
# Deprecation date: 2025-12-06
# Reason: Naming standardization to "CH-Annotator" convention
#
# DO NOT USE THIS FILE. Use the new location:
#   data/entity_annotation/ch_annotator-v1_7_0.yaml
# =============================================================================

# =============================================================================
# GLAM-NER: UNIFIED ENTITY ANNOTATION CONVENTION (DEPRECATED)
# =============================================================================
# Version: 1.7.0-unified
# Date: 2025-12-02
# Status: DEPRECATED - Replaced by ch_annotator-v1_7_0.yaml
# 
# PURPOSE: Complete convention for Named Entity Recognition, Property Extraction,
# Entity Resolution, Entity Linking, and Claim Validation across ALL text sources.
# Domain-agnostic rules that apply universally.
#
# This file is COMPLETE and SELF-CONTAINED. No external files required.
#
# CHANGELOG (1.7.0): DIGITAL HUMANITIES ONTOLOGY ALIGNMENT + DOCUMENT STRUCTURE
#
# PART A: HYPERNYM RESTRUCTURING (Breaking Changes)
# - BREAKING: Renamed BEING hypernym to AGENT (code: AGT) - includes non-human agents
#   following CIDOC-CRM E39_Actor (broader than E21_Person). Covers humans, animals,
#   AI agents, fictional characters, robots, and mythological beings (TEI CHR concept).
# - BREAKING: Renamed ORGANISATION hypernym to GROUP (code: GRP) - all collectives
#   following CIDOC-CRM E74_Group. Subcategories distinguish formal organizations
#   (org:FormalOrganization) from informal collectives (foaf:Group).
# - BREAKING: Split PLACE into TOPONYM (nominal place references) and GEOMETRY
#   (coordinate/shape data) following TEI/Pleiades/GeoSPARQL distinctions.
# - BREAKING: Restructured TEMPORAL_REFERENCE with TimeML/TIMEX3 type system:
#   * TMP.DAB: Datable (absolute timestamps, fully resolved to calendar)
#   * TMP.DRL: Deictic/Relative (require context for resolution)
#   * TMP.DUR: Durations (time:Duration, time:TemporalDuration)
#   * TMP.SET: Recurring/periodic times (time:Interval with frequency)
#   * TMP.RNG: Ranges (explicit start-end pairs)
# - BREAKING: Replaced nerd:Product mapping for TEXTUAL_REFERENCE with FRBR-based
#   model (frbr:Work, frbr:Expression, frbr:Manifestation, frbr:Item).
# - BREAKING: Added new hypernym ROLE (code: ROL) for positions, honorifics, titles,
#   and occupational designations following TEI <roleName> and PiCO role concepts.
# - Deprecated NERD mappings where they introduce inappropriate biases - NERD is
#   retained for cross-system interchange but is NOT authoritative for this convention.
# - Added Digital Humanities authority references: TEI P5 Chapter 14, TimeML/TIMEX3,
#   CIDOC-CRM 7.1.3, PiCO v1.0, FRBR, Pleiades, GeoSPARQL.
#
# PART B: SECTION 15 - DOCUMENT STRUCTURE AND NAMESPACE PATHS (New)
# - NEW: Added DOCUMENT_REGION (DOC) hypernym with comprehensive layout semantic ontology
#   for annotating document structure (headers, paragraphs, tables, sidebars, figures).
# - NEW: Primary content regions: DOC.HDR (6 heading levels), DOC.PAR, DOC.SEN, DOC.LST,
#   DOC.LIT, DOC.TBL (with TBL.HDR, TBL.BDY, TBL.ROW, TBL.CEL sub-components).
# - NEW: Supplementary content regions: DOC.SDB (Sidebar) with hyponyms DOC.SDB.MRG
#   (Marginalia with margin positions), DOC.SDB.IBX (Infobox), DOC.SDB.CLT (Callout).
# - NEW: Additional regions: DOC.CAP, DOC.FTN, DOC.FIG, DOC.NAV, DOC.PGN, DOC.BLK,
#   DOC.MTD (Metadata Block), DOC.ANN (Annotation Region).
# - NEW: Media collection region: DOC.GAL (Gallery) for image carousels, plate sections,
#   portfolio showcases - with gallery_types and item nesting.
# - NEW: Cartographic region: DOC.MAP (Map) for geographic visualizations with map_types
#   (reference, thematic, historical, site_plan, route, cadastral, interactive) and
#   map_components (MAP.BAS base layer, MAP.OVL overlay, MAP.MRK markers, MAP.SHP shapes,
#   MAP.LEG legend, MAP.CTL controls).
# - NEW: Audiovisual regions: DOC.AUD (Audio) with audio_types and audio_segments
#   (AUD.SPK speaker, AUD.MUS music, AUD.SIL silence, AUD.SFX effects); DOC.VID (Video)
#   with video_types and video_segments (VID.SCN scene, VID.SHT shot, VID.TTL title,
#   VID.CRD credits).
# - NEW: Embedded interactive region: DOC.EMB for iframes, widgets, IIIF viewers,
#   3D viewers, timelines, social embeds, data visualizations, virtual tours.
# - NEW: Navigation/reference regions promoted to full hypernyms: DOC.TOC (Table of
#   Contents with toc_types: main, LOF, LOT, abbreviations, maps, plates), DOC.IDX
#   (Index with index_types: name, place, subject, title, chronological - pre-annotated
#   entity sources).
# - NEW: Front matter regions: DOC.TTP (Title Page with components: TTL, STL, AUT, PUB,
#   DAT, PLC, EDT, IMP), DOC.DED (Dedication/Epigraph), DOC.COL (Colophon with components:
#   PRN printer, DAT date, PLC place, TYP typography, PAP paper, CPY copyright, EDN edition).
# - NEW: Back matter regions: DOC.BIB (Bibliography with entry components: AUT, TTL, DAT,
#   PUB, PLC), DOC.APP (Appendix), DOC.GLO (Glossary with components: TRM, DEF, SYN, REL).
# - NEW: Commercial/branding regions: DOC.ADV (Advertisement with types: display, classified,
#   trade_listing, prospectus, patent_medicine, auction_notice, legal_notice), DOC.LOG
#   (Logo with types: publisher_logo, printer_device, masthead, watermark, seal, coat_of_arms,
#   colophon_mark, ex_libris).
# - NEW: Semantic role enumeration expanded: PRIM, SUPP, NAV, STRC, REF, VIS, AV, INT, META,
#   SPAT, FRNT (front matter), BACK (back matter), PARA (paratextual), COMM (commercial),
#   LEX (lexical/definitions).
# - NEW: Two-layer nested provenance model - layout claims vs entity claims with separate
#   confidence/provenance chains. Entity claims reference parent layout claims.
# - NEW: Format-agnostic path conventions for PAGE-XML, HTML, JSON, Markdown, EPUB, PDF, TEI.
# - NEW: Clustering strategies for grouping entities by document region.
# - Added namespaces: premis:, bibo:, csvw:, html:, frbr:, frbre:, geosparql:, sf:,
#   timeml:, tei:, as: (ActivityStreams), skos: for Section 15 ontology mappings.
# - Added comprehensive LinkML mapping variants (class_uri, exact_mappings, 
#   close_mappings, related_mappings, narrow_mappings, broad_mappings).
#
# CHANGELOG (1.6.3):
# - MAJOR: Expanded scope from NER-only to full extraction pipeline
# - Added Property Extraction Rules section (relationship detection)
# - Added Claim Validation Schema (LinkML-based provenance)
# - Added Entity Resolution and Linking Pipeline section
# - Added LinkML namespace and schema references (https://linkml.io/)
# - Added temporal property extraction patterns (founding dates, etc.)
# - All claims now require 5-component provenance:
#   1. namespace (ontology prefix)
#   2. path (xpath/jsonpath)
#   3. timestamp (ISO 8601)
#   4. agent (extraction model)
#   5. context_convention (this file version)
# - Aligned with WebClaim.yaml LinkML schema for claim instances
#
# CHANGELOG (1.6.2):
# - Added NIF (NLP Interchange Format) vocabulary for annotation interchange
# - Added NERD ontology core class mappings for cross-system interoperability
# - Added W3C Web Annotation Data Model for text span provenance
# - Added itsrdf namespace for entity linking (ITS 2.0)
# - New section: NIF/NERD/OA Integration Patterns
# - Enhanced output format with NIF-compliant properties
#
# CHANGELOG (1.6.1):
# - Added W3C Org Ontology mappings (org:FormalOrganization, org:OrganizationalUnit)
# - Added RegOrg vocabulary mappings (rov:RegisteredOrganization, rov:legalName)
# - Added new ORG.UNT subcategory for organizational units
# - Enhanced BEING.STF with org:Membership, org:Role, org:Post patterns
# - Added foaf:OnlineAccount for TXT.SOC social media handles
# - Added Europeana Data Model (EDM) mappings for THING subcategories
# - Expanded relationships section with 14 new org:, rov:, foaf: predicates
# =============================================================================

convention:
  name: "GLAM-NER Unified Entity Annotation Convention"
  version: "1.7.0-unified"
  date: "2025-12-02"
  status: "production"
  
  description: |
    A unified, domain-agnostic convention for the complete entity extraction
    pipeline: Named Entity Recognition (NER), Property Extraction, Entity
    Resolution, Entity Linking, and Claim Validation.
    
    This convention applies universally to any text source: historical
    manuscripts, modern websites, archival documents, or digital platforms.
    
    The convention identifies 9 hypernym entity types (expanded from 8 in v1.6.x)
    with ontology mappings to CIDOC-CRM, RiC-O, Schema.org, PiCO, W3C Org,
    RegOrg, FOAF, vCard, EDM, OWL-Time, GeoSPARQL, FRBR, TEI, and NIF 2.0.
    
    CRITICAL DESIGN PRINCIPLES (v1.7.0):
    1. AGENT vs PERSON: "Being" is too narrow - AGENT covers humans, animals, AI,
       fictional characters, robots. Uses CIDOC-CRM E39_Actor as primary class.
    2. TOPONYM vs GEOMETRY: Place names (nominal references) are distinct from
       coordinate data (geometric representations). Never conflate these.
    3. GROUP vs ORGANISATION: "Organisation" implies formal structure - GROUP
       covers all collectives from informal bands to registered corporations.
    4. TEMPORAL distinctions: Absolute timestamps, relative expressions, durations,
       and recurring periods have fundamentally different semantics (TimeML/TIMEX3).
    5. ROLE is distinct from AGENT: "Director" is a role; "Dr. Jan de Wit" is an
       agent. Roles can be filled by different agents over time.
    6. TEXTUAL REFERENCE is not PRODUCT: Documents have Work/Expression/Manifestation
       distinctions (FRBR) that product-centric models miss.
    
    All extracted claims MUST include verifiable provenance following the LinkML
    schema at https://linkml.io/ and this project's WebClaim.yaml schema.
        
  digital_humanities_authorities:
    - name: "TEI P5 Guidelines, Chapter 14: Names, Dates, People, Places"
      url: "https://tei-c.org/release/doc/tei-p5-doc/en/html/ND.html"
      note: "Authoritative for persName, placeName, roleName, orgName distinctions"
    - name: "TimeML/TIMEX3 Specification"
      url: "https://timeml.github.io/site/timebank/"
      note: "Authoritative for temporal expression annotation types"
    - name: "CIDOC-CRM 7.1.3"
      url: "https://cidoc-crm.org/Version/version-7.1.3"
      note: "Authoritative for heritage domain entity classes"
    - name: "PiCO (Persons in Context) Ontology"
      url: "https://personsincontext.org/"
      note: "Authoritative for person observation/reconstruction pattern"
    - name: "FRBR (Functional Requirements for Bibliographic Records)"
      url: "https://www.ifla.org/publications/functional-requirements-for-bibliographic-records"
      note: "Authoritative for Work/Expression/Manifestation/Item distinctions"
    - name: "Pleiades Gazetteer"
      url: "https://pleiades.stoa.org/"
      note: "Model for place/name/location separation in historical geography"
    - name: "GeoSPARQL"
      url: "https://www.ogc.org/standard/geosparql/"
      note: "Authoritative for spatial geometry representation"
  
  nerd_deprecation_note: |
    NERD (Named Entity Recognition and Disambiguation) ontology mappings are
    RETAINED for cross-system NLP tool interchange but are NOT AUTHORITATIVE
    for this convention. NERD has biases toward modern web/journalism contexts:
    - nerd:Person excludes non-human agents (too narrow)
    - nerd:Location conflates toponyms with geometry (imprecise)
    - nerd:Organization excludes informal collectives (too narrow)
    - nerd:Time conflates absolute/relative/duration (imprecise)
    - nerd:Product misrepresents textual references (wrong semantics)
    - NERD lacks role/title/honorific class (missing category)
    
    Use NERD classes only for NIF export to downstream NLP tools that expect
    NERD vocabulary. Internal processing uses CIDOC-CRM, TEI, and TimeML classes.    
  
  scope:
    - entity_recognition: "Detect and classify named entities in text"
    - property_extraction: "Extract relationships and attributes between entities"
    - entity_resolution: "Disambiguate and merge entity mentions"
    - entity_linking: "Link entities to knowledge bases (Wikidata, etc.)"
    - claim_validation: "Verify and track provenance for all claims"
  
  linkml_reference:
    specification: "https://linkml.io/linkml/"
    version: "1.8.x"
    purpose: |
      LinkML (Linked Data Modeling Language) is used for:
      - Formal schema definitions (classes, slots, enums)
      - Instance validation
      - Multi-format serialization (YAML, JSON, RDF, etc.)
      - Type-safe claim structures      

# =============================================================================
# SECTION 1: ONTOLOGY NAMESPACES
# =============================================================================
# All namespace prefixes used in ontology mappings throughout this convention.

ontology:
  namespaces:
    # Core Heritage/Cultural Ontologies
    rico: "https://www.ica.org/standards/RiC/ontology#"
    crm: "http://www.cidoc-crm.org/cidoc-crm/"
    edm: "http://www.europeana.eu/schemas/edm/"
    
    # Web/General Ontologies  
    schema: "http://schema.org/"
    foaf: "http://xmlns.com/foaf/0.1/"
    
    # Organization Ontologies
    org: "http://www.w3.org/ns/org#"
    rov: "http://www.w3.org/ns/regorg#"
    
    # Person/Identity Ontologies
    picom: "https://personincontext.org/model/"
    pnv: "https://w3id.org/pnv#"
    
    # Contact Information
    vcard: "http://www.w3.org/2006/vcard/ns#"
    
    # Provenance
    prov: "http://www.w3.org/ns/prov#"
    
    # Vocabularies & Classification
    skos: "http://www.w3.org/2004/02/skos/core#"
    
    # Core RDF/OWL
    rdf: "http://www.w3.org/1999/02/22-rdf-syntax-ns#"
    rdfs: "http://www.w3.org/2000/01/rdf-schema#"
    xsd: "http://www.w3.org/2001/XMLSchema#"
    
    # Dublin Core
    dc: "http://purl.org/dc/elements/1.1/"
    dct: "http://purl.org/dc/terms/"
    
    # Temporal & Spatial
    time: "http://www.w3.org/2006/time#"
    geo: "http://www.w3.org/2003/01/geo/wgs84_pos#"
    
    # Measurement
    ivoa: "http://www.ivoa.net/rdf/UCD1+#"
    
    # Project-specific
    glam: "https://w3id.org/glam/"
    
    # NLP Interchange Format (NIF 2.0)
    nif: "http://persistence.uni-leipzig.org/nlp2rdf/ontologies/nif-core#"
    
    # Named Entity Recognition and Disambiguation (NERD)
    nerd: "http://nerd.eurecom.fr/ontology#"
    
    # W3C Web Annotation Data Model
    oa: "http://www.w3.org/ns/oa#"
    
    # Internationalization Tag Set (ITS 2.0) - Entity Linking
    itsrdf: "http://www.w3.org/2005/11/its/rdf#"
    
    # LinkML Schema Language
    linkml: "https://w3id.org/linkml/"
    
    # Heritage Custodian Ontology (project-specific LinkML schemas)
    hc: "https://nde.nl/ontology/hc/"
    
    # Provenance Authoring and Versioning
    pav: "http://purl.org/pav/"
    
    # Document Structure & Layout (Section 15)
    premis: "http://www.loc.gov/premis/rdf/v3/"
    bibo: "http://purl.org/ontology/bibo/"
    csvw: "http://www.w3.org/ns/csvw#"
    html: "http://www.w3.org/1999/xhtml/"
    
    # Bibliographic/FRBR (for TEXTUAL_REFERENCE)
    frbr: "http://purl.org/vocab/frbr/core#"
    frbre: "http://purl.org/vocab/frbr/extended#"
    
    # Spatial/Geometry (for GEOMETRY hypernym)
    geosparql: "http://www.opengis.net/ont/geosparql#"
    sf: "http://www.opengis.net/ont/sf#"
    
    # Temporal (TimeML/TIMEX3 extensions)
    timeml: "http://timeml.org/timeml/"
    
    # TEI (for CHR character model, roleName, etc.)
    tei: "http://www.tei-c.org/ns/1.0/"
    
    # ActivityStreams (for collections, media activities)
    as: "https://www.w3.org/ns/activitystreams#"

# =============================================================================
# SECTION 2: ENTITY TYPE DEFINITIONS
# =============================================================================
# Complete definitions for all 8 entity types with subcategories,
# inclusion/exclusion rules, and ontology mappings.

entity_types:

  # ---------------------------------------------------------------------------
  # BEING - Persons, animals, mythological figures
  # ---------------------------------------------------------------------------
  # ---------------------------------------------------------------------------
  # AGENT - Entities capable of intentional action (CIDOC-CRM E39_Actor)
  # ---------------------------------------------------------------------------
  # BREAKING CHANGE v1.7.0: Renamed from BEING (BEI) to AGENT (AGT)
  # Rationale: "Being" implies human-centric ontology. CIDOC-CRM E39_Actor is
  # the proper hypernym for ALL entities capable of intentional action:
  # humans, animals, AI systems, fictional characters, collectives, and robots.
  # ---------------------------------------------------------------------------
  AGENT:
    code: "AGT"
    definition: |
      Entities capable of intentional action. Includes humans (historical and
      contemporary), animals, AI systems, fictional characters, mythological
      figures, and collectives. The defining characteristic is AGENCY - the
      capacity to act, make decisions, and bear responsibility.
      
      This is the broadest actor class, encompassing all entities that can:
      - Perform intentional actions (create, destroy, transfer, modify)
      - Hold beliefs, desires, or goals
      - Bear moral or legal responsibility
      - Be attributed authorship or causation      
    
    design_rationale: |
      CIDOC-CRM E39_Actor ("a persistent item that has the potential to perform
      intentional actions") is the correct hypernym. The former "BEING" label
      was anthropocentric, excluding valid actors like:
      - AI systems creating art or making curatorial decisions
      - Named animals with documented agency (working animals, famous pets)
      - Fictional characters who are subjects of scholarly study
      - Robots performing heritage conservation tasks
      
      TEI P5 provides the Character (roleName) model for fictional entities.
      FOAF provides Agent as a superclass of Person and Group.      
    
    ontology_mappings:
      primary_class: "crm:E39_Actor"
      primary_class_definition: |
        CIDOC-CRM E39 Actor: "This class comprises people, either combineely as
        combinees of groups or combinees of groups. A gathering of combinees of
        E21 Persons becomes an instance of E74 Group when it exhibits collective
        agency, that is, it can perform actions as a unit."        
      alternative_classes:
        - "foaf:Agent"
        - "schema:Thing"  # schema:Person and schema:Organization are subclasses
        - "prov:Agent"
      linkml_mapping:
        class_uri: "crm:E39_Actor"
        exact_mappings:
          - "foaf:Agent"
          - "prov:Agent"
        close_mappings:
          - "schema:Person"  # More specific, human-only
        related_mappings:
          - "dcterms:Agent"
      nerd_class: "nerd:Person"
      nerd_deprecation_note: |
        DEPRECATED: NERD's Person class is too narrow. NERD was designed for
        news/journalism NER where non-human agents are rare. For Digital
        Humanities, CIDOC-CRM E39_Actor is authoritative.
        Retain NERD mapping ONLY for NLP pipeline interchange, NOT as semantic authority.        
      pico_class: "picom:PersonObservation"
      pico_note: |
        In PiCO, textual mentions create PersonObservation instances linked
        to reconstructed Person entities via picom:isObservationOf. This
        observation/reconstruction pattern applies to ALL agent subcategories.        
    
    subcategories:
      # ----- HUMAN AGENTS -----
      PERSON:
        code: "AGT.PER"
        definition: "Individual human beings, historical or contemporary"
        examples:
          - "Rembrandt van Rijn"
          - "Queen Beatrix"
          - "Jan de Bakker"
          - "Marie Curie"
        ontology_class: "crm:E21_Person"
        linkml_mapping:
          class_uri: "crm:E21_Person"
          exact_mappings:
            - "foaf:Person"
            - "schema:Person"
            - "rico:Person"
      
      STAFF:
        code: "AGT.STF"
        definition: "Personnel of heritage institutions in professional roles"
        examples:
          - "Dr. Maria van den Berg, Director"
          - "Jan Pietersen, Curator of Prints"
          - "Chief Archivist Emma de Vries"
        ontology_class: "picom:PersonObservation"
        org_ontology_mapping:
          membership: "org:Membership"
          role: "org:Role"
          post: "org:Post"
        note: |
          Links to institution via org:memberOf or org:holds (for Posts).
          The org:Membership class represents the n-ary relationship between
          an Agent, an Organization, and a Role. Use org:Post when the position
          exists independently of the person filling it.
          
          For the ROLE itself (e.g., "Director", "Curator"), see ROLE hypernym.          
      
      # ----- COLLECTIVE AGENTS -----
      COLLECTIVE:
        code: "AGT.COL"
        definition: |
          Named collectives of agents acting as a unit but WITHOUT formal
          organizational structure. For formal organizations, use GROUP hypernym.          
        examples:
          - "The Dutch Masters"
          - "The Impressionists"
          - "The Founding Fathers"
          - "Anonymous (hacker collective)"
        ontology_class: "crm:E74_Group"
        alternative_classes:
          - "foaf:Group"
        note: |
          Collectives exhibit collective agency but lack:
          - Legal personality
          - Formal membership rules
          - Organizational hierarchy
          
          For formal organizations (museums, companies), use GROUP hypernym.
          For informal project collaborations, use org:OrganizationalCollaboration.          
      
      # ----- FICTIONAL/MYTHOLOGICAL AGENTS -----
      FICTIONAL:
        code: "AGT.FIC"
        definition: |
          Characters from fiction, mythology, legend, or religious traditions
          who are subjects of scholarly study or cultural analysis.          
        examples:
          - "Sherlock Holmes"
          - "Harry Potter"
          - "Hamlet"
          - "Don Quixote"
        ontology_class: "crm:E21_Person"
        alternative_classes:
          - "tei:character"  # TEI P5 character element
        linkml_mapping:
          class_uri: "crm:E21_Person"
          exact_mappings: []
          close_mappings:
            - "schema:Person"
          note: "Use crm:P2_has_type with value 'fictional' to distinguish"
        tei_note: |
          TEI P5 uses <person> with @role="fictional" or nests within
          <listPerson type="fictional">. The <character> element (from
          TEI Drama module) is more specific for dramatic personae.          
      
      MYTHOLOGICAL:
        code: "AGT.MYT"
        definition: |
          Gods, deities, legendary figures, and supernatural beings from
          religious or mythological traditions.          
        examples:
          - "Apollo"
          - "Thor"
          - "Vishnu"
          - "Anansi"
          - "King Arthur"
          - "Siegfried"
        ontology_class: "crm:E21_Person"
        note: |
          Use crm:P2_has_type to indicate mythological/divine status.
          Mythological figures may have historical cult/worship data
          (temples, festivals) even though the entity is non-physical.          
      
      # ----- NON-HUMAN AGENTS -----
      ANIMAL:
        code: "AGT.ANI"
        definition: |
          Named individual animals with documented agency or cultural
          significance. NOT species names (use THING.TAX for taxonomy).          
        examples:
          - "Dolly the sheep (first cloned mammal)"
          - "Jumbo the elephant"
          - "Hachiko (famous loyal dog)"
          - "Wojtek the soldier bear"
          - "Paul the Octopus (World Cup predictor)"
        ontology_class: "crm:E39_Actor"
        note: |
          Animals qualify as agents when they:
          - Have individual names (not just species)
          - Performed documented actions
          - Have cultural/historical significance
          
          Generic animal mentions ("a cat", "the horses") are NOT agents.          
      
      ARTIFICIAL:
        code: "AGT.ART"
        definition: |
          Artificial agents: AI systems, robots, software agents, and
          automated systems capable of autonomous decision-making or action.          
        examples:
          - "DALL-E (AI image generator)"
          - "AlphaGo (game-playing AI)"
          - "Sophia (humanoid robot)"
          - "Watson (IBM's AI system)"
          - "GPT-4 (language model)"
        ontology_class: "crm:E39_Actor"
        alternative_classes:
          - "prov:SoftwareAgent"
        linkml_mapping:
          class_uri: "crm:E39_Actor"
          exact_mappings:
            - "prov:SoftwareAgent"
          note: |
            PROV-O defines prov:SoftwareAgent as "A software agent is running
            software." Use when the AI/robot is the proximate cause of an action,
            distinct from the human programmers or operators.            
        note: |
          Artificial agents are increasingly relevant for heritage:
          - AI systems making curatorial decisions
          - Robots performing conservation tasks
          - Automated digitization systems
          - AI-generated art and authorship questions
          
          Attribution of agency to AI is context-dependent and evolving.          
    
    inclusion_rules:
      - id: "AGT_INC001"
        rule: "Tag agent names even when only partial name appears"
        examples:
          - "Rembrandt (given name only)"
          - "Van Gogh (surname only)"
          - "GPT (abbreviated AI name)"
      
      - id: "AGT_INC002"
        rule: "Tag agents identified by title + name"
        examples:
          - "Professor Einstein"
          - "Dr. Curie"
          - "King Willem-Alexander"
      
      - id: "AGT_INC003"
        rule: "Tag staff members with their institutional context"
        examples:
          - "Director Jan de Wit"
          - "Curator of Medieval Art at the Rijksmuseum"
      
      - id: "AGT_INC004"
        rule: "Tag named collectives acting as unified agents"
        examples:
          - "The Impressionists"
          - "Anonymous"
          - "The Beatles"
      
      - id: "AGT_INC005"
        rule: "Tag fictional/mythological characters when subjects of analysis"
        examples:
          - "Hamlet's soliloquy"
          - "representations of Apollo"
          - "Harry Potter merchandise"
      
      - id: "AGT_INC006"
        rule: "Tag AI/robot agents when attributed with actions"
        examples:
          - "art created by DALL-E"
          - "AlphaGo defeated Lee Sedol"
    
    exclusion_rules:
      - id: "AGT_EXC001"
        rule: "Do NOT tag generic role descriptions without names"
        examples:
          - "the curator (generic)"
          - "a visitor (generic)"
          - "staff members (generic plural)"
        note: "For role terms themselves, see ROLE hypernym"
      
      - id: "AGT_EXC002"
        rule: "Do NOT tag pronouns"
        examples:
          - "he"
          - "she"
          - "they"
          - "it"
      
      - id: "AGT_EXC003"
        rule: "Do NOT tag species names (use THING.TAX instead)"
        examples:
          - "elephants (species, not individual)"
          - "Homo sapiens"
          - "cats (generic)"
      
      - id: "AGT_EXC004"
        rule: "Do NOT tag tools or software without agency attribution"
        examples:
          - "Photoshop (tool, not agent)"
          - "the database (system, not agent)"
        note: "Only tag AI when it is the attributed actor, not just a tool used"

  # ---------------------------------------------------------------------------
  # TOPONYM - Named place references (nominal, not geometric)
  # ---------------------------------------------------------------------------
  # BREAKING CHANGE v1.7.0: Renamed from PLACE (PLC) to TOPONYM (TOP)
  # Rationale: Distinguish between:
  #   - TOPONYM: Nominal place names ("Amsterdam", "the Alps") - textual references
  #   - GEOMETRY: Coordinates, polygons, spatial extents - see GEOMETRY hypernym
  # 
  # This follows the Pleiades model: Place = conceptual entity, Location = geometry.
  # TEI P5: <placeName> vs. <geo> elements.
  # GeoSPARQL: geo:Feature vs. geo:Geometry.
  # ---------------------------------------------------------------------------
  TOPONYM:
    code: "TOP"
    definition: |
      Named references to places in text. Toponyms are NOMINAL - they are
      linguistic labels for places, not the places themselves. The same place
      may have multiple toponyms (historical names, variant spellings, exonyms).
      
      Key distinction:
      - TOPONYM: "Amsterdam", "Constantinople", "the Netherlands" (names in text)
      - GEOMETRY: "52.3676° N, 4.9041° E" (coordinates, see GEOMETRY hypernym)
      
      A toponym can reference:
      - A persistent place (conceptual entity with temporal extent)
      - A location (specific geometry at a point in time)
      - An uncertain or legendary place (may lack precise geometry)      
    
    design_rationale: |
      The Pleiades gazetteer model distinguishes:
      - Place: A conceptual geographic entity with persistent identity
      - Name: A toponym (linguistic label) used for that place
      - Location: A geometry (coordinates) for that place at a specific time
      
      TEI P5 follows similar logic:
      - <placeName>: Nominal reference to a place
      - <geo>: Geographic coordinates (separate element)
      
      This separation is essential for:
      - Historical places with unknown geometry ("Atlantis", "El Dorado")
      - Places with changing geometry (coastlines, borders)
      - Places with multiple names over time (Constantinople→Istanbul)
      - Linking textual mentions to gazetteers (GeoNames, Pleiades, Wikidata)      
    
    ontology_mappings:
      primary_class: "crm:E53_Place"
      primary_class_definition: |
        CIDOC-CRM E53 Place: "This class comprises extents in space, in
        particular on the surface of the earth, in the pure sense of physics:
        independent from temporal phenomena and matter."        
      alternative_classes:
        - "schema:Place"
        - "rico:Place"
        - "edm:Place"
        - "pleiades:Place"
      linkml_mapping:
        class_uri: "crm:E53_Place"
        exact_mappings:
          - "edm:Place"
          - "rico:Place"
        close_mappings:
          - "schema:Place"
        related_mappings:
          - "gn:Feature"  # GeoNames
      nerd_class: "nerd:Location"
      nerd_deprecation_note: |
        DEPRECATED: NERD's Location class conflates toponyms (names) with
        geometry (coordinates). For Digital Humanities, use crm:E53_Place for
        conceptual places and geo:Geometry for spatial data.
        Retain NERD mapping ONLY for NLP pipeline interchange.        
      tei_mapping:
        element: "placeName"
        attributes:
          ref: "URI reference to gazetteer entry"
          type: "settlement|region|country|address|building|natural"
          cert: "high|medium|low (certainty of identification)"
      note: |
        EDM Place is equivalent to CIDOC-CRM E53_Place and is used in
        Europeana cultural heritage contexts. Pleiades provides the
        Place/Name/Location model for historical geography.        
    
    subcategories:
      # ----- ADMINISTRATIVE PLACES -----
      SETTLEMENT:
        code: "TOP.SET"
        definition: "Cities, towns, villages, and other populated places"
        examples:
          - "Amsterdam"
          - "New York City"
          - "the village of Giethoorn"
          - "古都京都 (Kyoto)"
        ontology_class: "schema:City"
        alternative_classes:
          - "gn:P.PPL"  # GeoNames populated place
        linkml_mapping:
          class_uri: "crm:E53_Place"
          close_mappings:
            - "schema:City"
            - "gn:P.PPL"
      
      REGION:
        code: "TOP.REG"
        definition: "Provinces, states, counties, and administrative regions"
        examples:
          - "North Holland"
          - "Bavaria"
          - "California"
          - "Île-de-France"
        ontology_class: "schema:AdministrativeArea"
        alternative_classes:
          - "gn:A.ADM1"  # GeoNames first-order admin division
      
      COUNTRY:
        code: "TOP.CTY"
        definition: "Nations and sovereign states (modern and historical)"
        examples:
          - "The Netherlands"
          - "France"
          - "Japan"
          - "the Dutch Republic (historical)"
        ontology_class: "schema:Country"
        note: |
          For historical polities (kingdoms, empires, republics), use
          crm:P2_has_type to indicate temporal status. Link to Wikidata
          for historical state succession chains.          
      
      # ----- PHYSICAL STRUCTURE PLACES -----
      ADDRESS:
        code: "TOP.ADR"
        definition: "Street addresses and postal locations"
        examples:
          - "Museumstraat 1, Amsterdam"
          - "1600 Pennsylvania Avenue"
          - "Postbus 74888, 1070 DN Amsterdam"
        ontology_class: "schema:PostalAddress"
        vcard_mapping:
          class: "vcard:Address"
          properties:
            street: "vcard:street-address"
            locality: "vcard:locality"
            region: "vcard:region"
            postal_code: "vcard:postal-code"
            country: "vcard:country-name"
        note: |
          Addresses are composite toponyms containing multiple components.
          Parse into structured vCard properties when possible.          
      
      INSTADDR:
        code: "TOP.IAD"
        definition: "Full institutional addresses including building names"
        examples:
          - "Rijksmuseum, Museumstraat 1, 1071 XX Amsterdam"
          - "British Museum, Great Russell St, London WC1B 3DG"
        ontology_class: "schema:PostalAddress"
        org_mapping: "org:Site"
        note: |
          Links to GROUP hypernym via schema:address or org:hasSite.
          org:Site represents an office or premise at which the organization
          is located - use for physical institutional locations.          
      
      BUILDING:
        code: "TOP.BLD"
        definition: "Named buildings, monuments, and architectural structures"
        examples:
          - "the Rijksmuseum building"
          - "Anne Frank House"
          - "Palace of Versailles"
          - "the Parthenon"
        ontology_class: "crm:E18_Physical_Thing"
        alternative_classes:
          - "edm:PhysicalThing"
          - "schema:LandmarksOrHistoricalBuildings"
        note: |
          Buildings are physical things (E18) that occupy places (E53).
          The building-as-place uses crm:P53_has_former_or_current_location.          
      
      # ----- NATURAL FEATURE PLACES -----
      NATURAL:
        code: "TOP.NAT"
        definition: "Natural geographic features: mountains, rivers, lakes, etc."
        examples:
          - "the Alps"
          - "Amazon River"
          - "Mount Fuji"
          - "Lake Baikal"
          - "the Sahara Desert"
        ontology_class: "crm:E53_Place"
        alternative_classes:
          - "gn:T"  # GeoNames terrain features
          - "gn:H"  # GeoNames hydrographic features
      
      # ----- TEMPORAL/UNCERTAIN PLACES -----
      HISTORICAL:
        code: "TOP.HIS"
        definition: |
          Historical toponyms: places that no longer exist, have changed
          names, or have uncertain modern equivalents.          
        examples:
          - "Constantinople (→ Istanbul)"
          - "Batavia (→ Jakarta)"
          - "New Amsterdam (→ New York)"
          - "Babylon"
          - "Tenochtitlan"
        ontology_class: "crm:E53_Place"
        pleiades_note: |
          Pleiades is the authoritative gazetteer for ancient world places.
          Use pleiades:Place identifiers for Greco-Roman and ancient Near
          Eastern toponyms. Link via @ref in TEI.          
        note: |
          Use crm:P2_has_type to indicate historical status.
          Link historical toponyms to modern equivalents via owl:sameAs
          or skos:closeMatch when identity is certain.          
      
      LEGENDARY:
        code: "TOP.LEG"
        definition: |
          Legendary, mythological, or fictional places. These may have
          cultural significance but lack verifiable geometry.          
        examples:
          - "Atlantis"
          - "El Dorado"
          - "Avalon"
          - "Middle-earth"
          - "Narnia"
        ontology_class: "crm:E53_Place"
        tei_mapping:
          element: "placeName"
          attributes:
            type: "mythological"
            cert: "low"
        note: |
          Tag legendary places when they are subjects of scholarly analysis
          (archaeology, literary studies, art history). Use crm:P2_has_type
          to indicate legendary/fictional status.          
    
    inclusion_rules:
      - id: "TOP_INC001"
        rule: "Tag place names even with directional or temporal modifiers"
        examples:
          - "northern France"
          - "East Berlin"
          - "southern Netherlands"
          - "medieval Paris"
      
      - id: "TOP_INC002"
        rule: "Tag complete addresses as single entities"
        examples:
          - "Prinsengracht 263, 1016 GV Amsterdam"
      
      - id: "TOP_INC003"
        rule: "Tag historical toponyms with their historical form"
        examples:
          - "Batavia (not Jakarta, unless both appear)"
          - "Constantinople (not Istanbul)"
      
      - id: "TOP_INC004"
        rule: "Tag exonyms (foreign names for places)"
        examples:
          - "The Hague (English exonym for Den Haag)"
          - "Florence (English for Firenze)"
          - "Cologne (English for Köln)"
      
      - id: "TOP_INC005"
        rule: "Tag legendary places when subjects of scholarly analysis"
        examples:
          - "the search for Atlantis"
          - "representations of Avalon in medieval art"
    
    exclusion_rules:
      - id: "TOP_EXC001"
        rule: "Do NOT tag generic spatial references"
        examples:
          - "here, there, nearby"
          - "the city (without name)"
          - "the museum (use GROUP hypernym)"
      
      - id: "TOP_EXC002"
        rule: "Do NOT tag directional words alone"
        examples:
          - "north"
          - "south"
          - "east"
          - "west"
      
      - id: "TOP_EXC003"
        rule: "Do NOT tag room names within buildings (unless historically significant)"
        examples:
          - "Room 5"
          - "the lobby"
        note: "Exception: 'the Sistine Chapel' (historically significant)"
      
      - id: "TOP_EXC004"
        rule: "Do NOT tag coordinates or geometries (use GEOMETRY hypernym)"
        examples:
          - "52.3676° N, 4.9041° E (use GEO.PNT)"
          - "bounding box: 52.0, 4.5, 53.0, 5.0 (use GEO.BOX)"

  # ---------------------------------------------------------------------------
  # GEOMETRY - Spatial coordinates and geometric representations
  # ---------------------------------------------------------------------------
  # NEW in v1.7.0: Separated from PLACE to distinguish nominal (toponyms)
  # from geometric (coordinates) representations of space.
  # Authority: GeoSPARQL, OGC Simple Features, ISO 19107 (geographic info)
  # ---------------------------------------------------------------------------
  GEOMETRY:
    code: "GEO"
    definition: |
      Geometric representations of spatial extent: coordinates, bounding boxes,
      polygons, and other spatial primitives. Geometries are MATHEMATICAL
      representations, distinct from nominal toponyms.
      
      Key distinction:
      - TOPONYM: "Amsterdam" (name in text)
      - GEOMETRY: "52.3676° N, 4.9041° E" (coordinates)
      
      Geometries may be:
      - Extracted from text (coordinate mentions)
      - Resolved from toponyms (geocoding)
      - Imported from GIS data (shapefiles, GeoJSON)      
    
    design_rationale: |
      GeoSPARQL (OGC standard) distinguishes:
      - geo:Feature: An entity with spatial extent (≈ crm:E53_Place)
      - geo:Geometry: A geometric representation of that extent
      
      A single Feature may have multiple Geometries:
      - Different precision levels
      - Different time periods (historical boundaries)
      - Different representations (point vs. polygon)
      
      This separation enables:
      - Linking toponyms to multiple coordinate systems
      - Representing uncertainty in historical geography
      - Integrating with GIS systems      
    
    ontology_mappings:
      primary_class: "geo:Geometry"
      primary_class_definition: |
        GeoSPARQL geo:Geometry: "A coherent set of direct positions in space.
        The basic geometries are Point, Curve, Surface, Solid."        
      alternative_classes:
        - "sf:Geometry"  # OGC Simple Features
        - "crm:E94_Space_Primitive"  # CIDOC-CRM spatial extension
      linkml_mapping:
        class_uri: "geo:Geometry"
        exact_mappings:
          - "sf:Geometry"
        related_mappings:
          - "schema:GeoCoordinates"
          - "schema:GeoShape"
      tei_mapping:
        element: "geo"
        note: |
          TEI P5 <geo> contains coordinates in decimal degrees (WGS84).
          Format: latitude,longitude or latitude longitude (space-separated).          
      note: |
        GeoSPARQL is the W3C/OGC standard for representing geographic
        information in RDF. Use geo:hasGeometry to link Features to Geometries.        
    
    subcategories:
      POINT:
        code: "GEO.PNT"
        definition: "A single coordinate point (latitude/longitude)"
        examples:
          - "52.3676° N, 4.9041° E"
          - "52.3676, 4.9041"
          - "lat: 52.3676, lon: 4.9041"
          - "N 52° 22' 3.36\", E 4° 54' 14.76\""
        ontology_class: "sf:Point"
        alternative_classes:
          - "schema:GeoCoordinates"
        geosparql_wkt: "POINT(4.9041 52.3676)"
        note: |
          Points may appear in various formats:
          - Decimal degrees (DD): 52.3676, 4.9041
          - Degrees minutes seconds (DMS): N 52° 22' 3.36"
          - Signed decimal: 52.3676, 4.9041 (N/E positive)
          
          Normalize to WGS84 decimal degrees for storage.          
      
      BOX:
        code: "GEO.BOX"
        definition: "A bounding box (minimum bounding rectangle)"
        examples:
          - "bounding box: 52.0, 4.5, 53.0, 5.0"
          - "extent: SW 52.0, 4.5 to NE 53.0, 5.0"
        ontology_class: "geo:Geometry"
        alternative_classes:
          - "schema:GeoShape"
        geosparql_wkt: "ENVELOPE(4.5, 5.0, 53.0, 52.0)"
        note: |
          Bounding boxes define rectangular extents.
          Format conventions vary; normalize to: minLat, minLon, maxLat, maxLon.          
      
      POLYGON:
        code: "GEO.PLY"
        definition: "A closed polygon (administrative boundary, parcel)"
        examples:
          - "POLYGON((4.8 52.3, 4.9 52.3, 4.9 52.4, 4.8 52.4, 4.8 52.3))"
          - "GeoJSON polygon for Noord-Holland province"
        ontology_class: "sf:Polygon"
        geosparql_wkt: "POLYGON((...))"
        note: |
          Polygons are typically imported from GIS data, not extracted from
          natural language text. May appear in technical documentation.          
      
      LINE:
        code: "GEO.LIN"
        definition: "A line or path (route, river course, boundary segment)"
        examples:
          - "the route from Amsterdam to Rotterdam"
          - "LINESTRING(4.9 52.4, 4.5 51.9)"
        ontology_class: "sf:LineString"
        note: |
          Lines appear in route descriptions or boundary definitions.
          Less common in heritage text; more typical in geographic data.          
    
    inclusion_rules:
      - id: "GEO_INC001"
        rule: "Tag coordinate mentions in any standard format"
        examples:
          - "located at 52.3676° N, 4.9041° E"
          - "coordinates: 52.3676, 4.9041"
          - "GPS: N 52° 22' 3.36\", E 4° 54' 14.76\""
      
      - id: "GEO_INC002"
        rule: "Tag WKT (Well-Known Text) geometry literals"
        examples:
          - "POINT(4.9041 52.3676)"
          - "POLYGON((...))"
      
      - id: "GEO_INC003"
        rule: "Tag GeoJSON references when geometry is inline"
        examples:
          - '{"type": "Point", "coordinates": [4.9041, 52.3676]}'
    
    exclusion_rules:
      - id: "GEO_EXC001"
        rule: "Do NOT tag place names (use TOPONYM hypernym)"
        examples:
          - "Amsterdam (use TOP.SET)"
          - "the Alps (use TOP.NAT)"
      
      - id: "GEO_EXC002"
        rule: "Do NOT tag references to external GIS files"
        examples:
          - "see shapefile boundaries.shp (external reference)"
          - "GeoJSON file: regions.geojson (external)"
        note: "Only tag inline geometry, not file references"

  # ---------------------------------------------------------------------------
  # GROUP - Collectives of agents (formal and informal)
  # ---------------------------------------------------------------------------
  # BREAKING CHANGE v1.7.0: Renamed from ORGANISATION (ORG) to GROUP (GRP)
  # Rationale: CIDOC-CRM E74_Group is the proper hypernym for ALL collectives:
  #   - Formal organizations (legal entities, corporations, governments)
  #   - Informal groups (movements, bands, collectives, families)
  #   - Historical entities (guilds, societies, courts)
  # 
  # NERD's "Organization" is too narrow - it implies formal legal structure.
  # E74_Group: "any gathering or organization of E39 Actors that acts collectively"
  # ---------------------------------------------------------------------------
  GROUP:
    code: "GRP"
    definition: |
      Any collective of agents acting together. Groups range from formal
      legal entities (corporations, governments, universities) to informal
      collectives (bands, movements, families, friend groups).
      
      The defining characteristic is COLLECTIVE AGENCY - the group can:
      - Perform actions attributed to the group as a unit
      - Hold collective identity distinct from individual members
      - Persist through changes in membership
      
      Groups are subclasses of AGENT (crm:E74_Group is a subclass of E39_Actor).      
    
    design_rationale: |
      CIDOC-CRM E74_Group: "This class comprises any gatherings or organizations
      of E21 Persons that act collectively or in a similar way due to any form
      of unifying relationship."
      
      This is broader than:
      - org:Organization (W3C Org) - implies formal structure
      - foaf:Organization - primarily corporate entities
      - schema:Organization - web-focused commercial entities
      
      E74_Group encompasses:
      - Formal organizations (org:FormalOrganization)
      - Informal groups (foaf:Group)
      - Historical entities (guilds, courts, movements)
      - Families and dynasties
      - Artistic/cultural movements without formal structure      
    
    ontology_mappings:
      primary_class: "crm:E74_Group"
      primary_class_definition: |
        CIDOC-CRM E74 Group: "This class comprises any gatherings or organizations
        of E21 Persons that act collectively or in a similar way due to any form
        of unifying relationship."        
      alternative_classes:
        - "rico:CorporateBody"  # For formal organizations
        - "foaf:Group"  # For informal groups
        - "org:Organization"  # W3C Org for structured organizations
      linkml_mapping:
        class_uri: "crm:E74_Group"
        exact_mappings:
          - "foaf:Group"
        close_mappings:
          - "org:Organization"
          - "schema:Organization"
        related_mappings:
          - "rico:CorporateBody"
      nerd_class: "nerd:Organization"
      nerd_deprecation_note: |
        DEPRECATED: NERD's Organization class implies formal legal structure,
        excluding informal groups, movements, and historical collectives. For
        Digital Humanities, CIDOC-CRM E74_Group is authoritative.
        Retain NERD mapping ONLY for NLP pipeline interchange.        
      org_ontology:
        formal_organization: "org:FormalOrganization"
        organizational_unit: "org:OrganizationalUnit"
        note: |
          W3C Org Ontology provides precise organizational modeling:
          - org:Organization: Any collection of people organized together
          - org:FormalOrganization: Recognized legal entity (company, charity, government)
          - org:OrganizationalUnit: Division or department within an organization
          
          Use org:FormalOrganization for GRP.HER, GRP.COR, GRP.GOV, GRP.EDU.
          Use foaf:Group or crm:E74_Group for GRP.INF.          
      rov_ontology:
        registered_organization: "rov:RegisteredOrganization"
        legal_name: "rov:legalName"
        org_type: "rov:orgType"
        org_status: "rov:orgStatus"
        registration: "rov:registration"
        note: |
          Registered Organization Vocabulary (RegOrg) for legal entities:
          - rov:RegisteredOrganization: Legal entity in a formal register
          - rov:legalName: Registered legal name (vs. trading/brand names)
          - Use ONLY for formal organizations with legal registration.          
    
    subcategories:
      # ----- FORMAL ORGANIZATIONS -----
      HERINST:
        code: "GRP.HER"
        definition: "Heritage institutions: museums, archives, libraries, galleries"
        examples:
          - "Rijksmuseum"
          - "National Archives of the Netherlands"
          - "British Library"
          - "Louvre Museum"
        ontology_class: "glam:HeritageCustodian"
        alternative_classes:
          - "org:FormalOrganization"
          - "rov:RegisteredOrganization"
        linkml_mapping:
          class_uri: "glam:HeritageCustodian"
          exact_mappings: []
          close_mappings:
            - "schema:Museum"
            - "schema:Library"
            - "schema:ArchiveOrganization"
        note: |
          Primary entity type for GLAM project extraction.
          Most heritage institutions are formal organizations with legal registration.
          Use rov:RegisteredOrganization when registration details (KvK, ISIL) are known.          
      
      PARBODY:
        code: "GRP.PAR"
        definition: "Parent or governing bodies of heritage institutions"
        examples:
          - "Ministry of Education, Culture and Science"
          - "Smithsonian Institution"
          - "City of Amsterdam"
        ontology_class: "rico:CorporateBody"
        alternative_classes:
          - "org:Organization"
        org_predicates:
          - "org:hasSubOrganization"
          - "org:hasUnit"
        note: "Links to HERINST via org:hasSubOrganization or org:hasUnit"
      
      UNIT:
        code: "GRP.UNT"
        definition: "Departments, divisions, or units within organizations"
        examples:
          - "Department of Prints and Drawings"
          - "Conservation Laboratory"
          - "Education Services"
          - "Digital Collections Unit"
        ontology_class: "org:OrganizationalUnit"
        org_predicates:
          - "org:unitOf"
          - "org:hasUnit"
        note: |
          Use for named internal divisions when they function as distinct entities.
          Links to parent via org:unitOf. Distinct from GRP_EXC003 generic mentions.          
      
      CORPORATION:
        code: "GRP.COR"
        definition: "Commercial companies and businesses"
        examples:
          - "Dutch East India Company (VOC)"
          - "Philips"
          - "Royal Dutch Shell"
          - "Google"
        ontology_class: "schema:Corporation"
        alternative_classes:
          - "org:FormalOrganization"
          - "rov:RegisteredOrganization"
        note: |
          Historical trading companies (VOC, WIC, East India Company) are
          GRP.COR even though defunct. Use crm:P2_has_type for historical status.          
      
      GOVERNMENT:
        code: "GRP.GOV"
        definition: "Government agencies, legislatures, and public bodies"
        examples:
          - "Dutch Parliament (Staten-Generaal)"
          - "European Commission"
          - "City Council of Amsterdam"
          - "US National Endowment for the Humanities"
        ontology_class: "schema:GovernmentOrganization"
        alternative_classes:
          - "org:FormalOrganization"
      
      EDUCATIONAL:
        code: "GRP.EDU"
        definition: "Universities, schools, and educational institutions"
        examples:
          - "University of Amsterdam"
          - "Leiden University"
          - "Harvard University"
          - "Royal Academy of Fine Arts"
        ontology_class: "schema:EducationalOrganization"
        alternative_classes:
          - "org:FormalOrganization"
      
      RELIGIOUS:
        code: "GRP.REL"
        definition: "Religious organizations, denominations, and congregations"
        examples:
          - "Roman Catholic Church"
          - "Westerkerk congregation"
          - "Buddhist Temple Foundation"
          - "Franciscan Order"
        ontology_class: "schema:ReligiousOrganization"
        note: |
          Religious organizations may also be heritage custodians (GRP.HER) when
          they maintain archives, libraries, or collections. Use multiple types.          
      
      ASSOCIATION:
        code: "GRP.ASS"
        definition: "Associations, societies, and membership organizations"
        examples:
          - "Royal Netherlands Academy of Arts and Sciences"
          - "Historical Society of Amsterdam"
          - "Friends of the Rijksmuseum"
          - "International Council of Museums (ICOM)"
        ontology_class: "org:FormalOrganization"
        alternative_classes:
          - "schema:Organization"
        note: |
          Associations are formal groups with membership structures but may
          not have commercial purposes. Distinguished from GRP.INF by having
          formal bylaws, officers, and registered status.          
      
      # ----- INFORMAL GROUPS (NEW in v1.7.0) -----
      INFORMAL:
        code: "GRP.INF"
        definition: |
          Informal collectives without legal structure: artistic movements,
          social groups, loose collaborations, families, dynasties.          
        examples:
          - "The Impressionists"
          - "De Stijl movement"
          - "Anonymous (hacker collective)"
          - "the Habsburg dynasty"
          - "the Medici family"
          - "the Beatles"
        ontology_class: "crm:E74_Group"
        alternative_classes:
          - "foaf:Group"
        linkml_mapping:
          class_uri: "crm:E74_Group"
          exact_mappings:
            - "foaf:Group"
        note: |
          Informal groups exhibit collective agency but lack:
          - Legal personality or registration
          - Formal membership rules
          - Organizational hierarchy
          
          Artistic movements, dynasties, bands, and collectives are GRP.INF.
          The "group" is often named retrospectively by historians/critics.          
      
      # ----- HISTORICAL GROUPS -----
      HISTORICAL:
        code: "GRP.HIS"
        definition: "Historical organizations that no longer exist or have transformed"
        examples:
          - "Dutch East India Company (dissolved 1799)"
          - "Guild of Saint Luke"
          - "Knights Templar"
          - "League of Nations"
          - "the Hanseatic League"
        ontology_class: "crm:E74_Group"
        note: |
          Use crm:P2_has_type to indicate historical/defunct status.
          Historical groups may have successor organizations - link via
          org:resultedFrom or crm:P17_was_motivated_by.          
    
    inclusion_rules:
      - id: "GRP_INC001"
        rule: "Tag groups with their full official or common names"
        examples:
          - "Rijksmuseum Amsterdam"
          - "The Metropolitan Museum of Art"
          - "De Stijl"
      
      - id: "GRP_INC002"
        rule: "Tag abbreviated group names and acronyms"
        examples:
          - "VOC (Dutch East India Company)"
          - "MoMA (Museum of Modern Art)"
          - "KNAW (Royal Netherlands Academy)"
      
      - id: "GRP_INC003"
        rule: "Tag historical groups even if defunct"
        examples:
          - "Dutch East India Company"
          - "League of Nations"
          - "Guild of Saint Luke"
      
      - id: "GRP_INC004"
        rule: "Tag parent bodies and governing organizations"
        examples:
          - "under the Ministry of Culture"
          - "part of the Smithsonian"
      
      - id: "GRP_INC005"
        rule: "Tag informal groups and movements with recognized names"
        examples:
          - "the Impressionists"
          - "De Stijl movement"
          - "the Habsburg dynasty"
      
      - id: "GRP_INC006"
        rule: "Tag families and dynasties when acting as collective agents"
        examples:
          - "the Medici family commissioned..."
          - "Habsburg patronage of the arts"
    
    exclusion_rules:
      - id: "GRP_EXC001"
        rule: "Do NOT tag generic group references"
        examples:
          - "the museum (without name)"
          - "our organization"
          - "the company"
          - "a group of artists"
      
      - id: "GRP_EXC002"
        rule: "Do NOT tag group types without names"
        examples:
          - "museums (generic plural)"
          - "archives (generic)"
          - "families (generic)"
      
      - id: "GRP_EXC003"
        rule: "Do NOT tag departments unless they function as named units"
        examples:
          - "the prints department (use only if named entity)"
          - "HR department (generic)"
      
      - id: "GRP_EXC004"
        rule: "Do NOT tag ad-hoc gatherings without persistent identity"
        examples:
          - "the crowd at the opening"
          - "visitors to the museum"
          - "the workshop participants"

  # ---------------------------------------------------------------------------
  # APPELLATION - Names and naming constructs (distinct from what they name)
  # ---------------------------------------------------------------------------
  # BREAKING CHANGE v1.7.0: Renamed from DENOMINATION (DEN) to APPELLATION (APP)
  # and restructured to separate:
  #   - APPELLATION: Names as linguistic constructs (titles, collection names)
  #   - ROLE: Social/occupational positions (see new ROLE hypernym below)
  # 
  # TEI P5 provides the authority: <name>, <title>, <roleName>, <occupation>
  # are distinct element types, not conflated.
  # ---------------------------------------------------------------------------
  APPELLATION:
    code: "APP"
    definition: |
      Naming constructs as linguistic entities, distinct from the things they
      name. Appellations are linguistic labels that identify entities:
      - Titles (artwork titles, document titles)
      - Collection names (archival fonds, named collections)
      - Exhibition names (named temporary displays)
      - Awards (named honors and prizes)
      - Structured person names (when analyzing name components)
      
      For ROLES (positions, occupations, honorifics), see the ROLE hypernym.      
    
    design_rationale: |
      CIDOC-CRM E41 Appellation: "This class comprises signs, either meaningful
      or not, or arrangements of signs following a specific syntax, that are
      used or can be used to refer to and identify a specific instance."
      
      Appellations are DISTINCT from what they name:
      - "The Night Watch" (appellation) vs. the painting (E22_Human-Made_Object)
      - "Rembrandt van Rijn" (appellation) vs. the person (E21_Person)
      
      TEI P5 uses:
      - <name> for any naming string
      - <persName> for person names (with sub-elements)
      - <placeName> for place names
      - <title> for work titles
      - <roleName> for role designations (see ROLE hypernym)      
    
    ontology_mappings:
      primary_class: "crm:E41_Appellation"
      primary_class_definition: |
        CIDOC-CRM E41 Appellation: "This class comprises signs, either meaningful
        or not, or arrangements of signs following a specific syntax, that are
        used or can be used to refer to and identify a specific instance."        
      alternative_classes:
        - "skos:prefLabel"
        - "rdfs:label"
        - "rico:Name"
      linkml_mapping:
        class_uri: "crm:E41_Appellation"
        exact_mappings: []
        close_mappings:
          - "skos:prefLabel"
          - "rdfs:label"
      nerd_class: null
      nerd_note: |
        NERD has no dedicated class for names/appellations. Appellation entities
        are linguistic constructs that name other entities. In NIF export, these
        become nif:Phrase instances without a specific NERD type.        
      pnv_class: "pnv:PersonName"
      pnv_note: |
        For person names, use PNV (Person Name Vocabulary) structured representation
        with components: pnv:givenName, pnv:surnamePrefix, pnv:baseSurname, pnv:patronym        
    
    subcategories:
      TITLE:
        code: "APP.TIT"
        definition: "Titles of artworks, books, documents, musical works"
        examples:
          - "The Night Watch"
          - "Girl with a Pearl Earring"
          - "On the Origin of Species"
          - "Symphony No. 9"
        ontology_class: "crm:E35_Title"
        linkml_mapping:
          class_uri: "crm:E35_Title"
        note: |
          Titles are appellations for Works (see WORK hypernym).
          The same work may have multiple titles (original, translated, popular).          
      
      COLLECTION:
        code: "APP.COL"
        definition: "Named collections and archival fonds"
        examples:
          - "the Golden Age Collection"
          - "Archief van de Familie Bicker"
          - "Dutch Masters Collection"
          - "Fonds Bicker"
        ontology_class: "rico:RecordSet"
        alternative_classes:
          - "crm:E78_Curated_Holding"
        note: |
          Collection names identify aggregations of objects.
          Use rico:RecordSet for archival fonds/series.
          Use crm:E78_Curated_Holding for museum collections.          
      
      EXHIBITION:
        code: "APP.EXH"
        definition: "Named exhibitions and displays"
        examples:
          - "Rembrandt and the Golden Age"
          - "Vermeer: Master of Light"
          - "The Art of Living (2023)"
        ontology_class: "crm:E7_Activity"
        alternative_classes:
          - "schema:ExhibitionEvent"
        note: |
          Exhibition names are appellations for events (E7_Activity).
          Include dates when analyzing exhibition history.          
      
      AWARD:
        code: "APP.AWD"
        definition: "Named prizes, honors, and awards"
        examples:
          - "Nobel Prize in Physics"
          - "Rijksmuseum Award"
          - "Order of Orange-Nassau"
          - "Pulitzer Prize"
        ontology_class: "crm:E55_Type"
        note: |
          Awards are types that can be bestowed. The bestowal event
          is an E7_Activity; the award type is the appellation.          
      
      PERSONNAME:
        code: "APP.PNM"
        definition: |
          Structured person name components, for detailed name analysis.
          Use when parsing name parts, not for simple person mentions.          
        examples:
          - "van Rijn (surname prefix + surname)"
          - "Rembrandt Harmenszoon (given name + patronym)"
          - "Dr. Maria van den Berg, PhD"
        ontology_class: "pnv:PersonName"
        pnv_components:
          givenName: "pnv:givenName"
          patronym: "pnv:patronym"
          surnamePrefix: "pnv:surnamePrefix"
          baseSurname: "pnv:baseSurname"
          honorificPrefix: "pnv:honorificPrefix (Dr., Prof.)"
          honorificSuffix: "pnv:honorificSuffix (PhD, Jr.)"
        note: |
          Use APP.PNM when analyzing name structure (genealogy, prosopography).
          For simple person mentions, use AGENT.PER.
          Honorifics in names are part of the appellation, not separate roles.          
    
    inclusion_rules:
      - id: "APP_INC001"
        rule: "Tag work titles in their original language when possible"
        examples:
          - "De Nachtwacht (Dutch)"
          - "The Night Watch (English translation)"
      
      - id: "APP_INC002"
        rule: "Tag collection names including archival fonds designations"
        examples:
          - "Fonds Bicker"
          - "Collection Fodor"
      
      - id: "APP_INC003"
        rule: "Tag exhibition titles with dates when given"
        examples:
          - "Vermeer (2023)"
          - "Rembrandt: The Late Works (2015)"
    
    exclusion_rules:
      - id: "APP_EXC001"
        rule: "Do NOT tag generic descriptive phrases"
        examples:
          - "the painting (not a title)"
          - "the exhibition (not named)"
      
      - id: "APP_EXC002"
        rule: "Do NOT double-tag: if tagged as AGENT/GROUP, do not also tag as APPELLATION"
        note: "Exception: when specifically analyzing name structure (APP.PNM)"
      
      - id: "APP_EXC003"
        rule: "Do NOT tag roles/occupations as appellations (use ROLE hypernym)"
        examples:
          - "Director (use ROLE.OCC)"
          - "Curator (use ROLE.OCC)"
          - "His Majesty (use ROLE.HON)"

  # ---------------------------------------------------------------------------
  # ROLE - Social positions, occupations, and honorifics (TEI roleName model)
  # ---------------------------------------------------------------------------
  # NEW in v1.7.0: Separated from DENOMINATION to properly model:
  #   - Occupational roles (curator, archivist, director)
  #   - Honorific titles (Your Majesty, Professor, Dr.)
  #   - Social positions (president, secretary, member)
  #   - Relational roles (father of, student of)
  # 
  # TEI P5: <roleName> is distinct from <persName> and <occupation>
  # W3C Org: org:Role and org:Post model organizational positions
  # ---------------------------------------------------------------------------
  ROLE:
    code: "ROL"
    definition: |
      Social, occupational, and ceremonial positions that agents hold or
      have held. Roles are POSITIONS, not the persons who fill them.
      
      Key distinction:
      - AGENT: The person ("Jan de Wit")
      - ROLE: The position ("Director", "Curator")
      - APPELLATION: The name as linguistic construct ("Dr. Jan de Wit")
      
      Roles include:
      - Occupations: Professional positions (curator, archivist, librarian)
      - Titles: Honorific designations (Professor, Dr., Sir)
      - Offices: Formal positions in organizations (Director, President)
      - Relational: Family/social relationships (father of, mentor to)      
    
    design_rationale: |
      TEI P5 distinguishes:
      - <roleName>: "contains a name component which indicates that the
        referent has a particular role or position in society"
      - <occupation>: "contains an informal description of a person's trade,
        profession, or occupation"
      
      W3C Org Ontology provides:
      - org:Role: A role within an organization
      - org:Post: A specific position that exists independently of holders
      - org:Membership: The n-ary relation of agent+organization+role
      
      CIDOC-CRM uses:
      - E55_Type for role classification
      - P14.1_in_the_role_of for role-in-activity
      
      Roles are NOT the same as:
      - AGENT (the person, not the position)
      - APPELLATION (the name, not the social function)
      - GROUP (the organization, not the position within it)      
    
    ontology_mappings:
      primary_class: "org:Role"
      primary_class_definition: |
        W3C Org Ontology org:Role: "Denotes a role that a Person or other Agent
        can take in an organization. Instances of this class describe the
        abstract role; to denote a specific instance of a person playing that
        role in a specific organization use a Membership."        
      alternative_classes:
        - "org:Post"  # Specific position (e.g., "Director of the Rijksmuseum")
        - "crm:E55_Type"  # CIDOC-CRM classification
        - "schema:Role"  # Schema.org
      linkml_mapping:
        class_uri: "org:Role"
        exact_mappings:
          - "schema:Role"
        close_mappings:
          - "crm:E55_Type"
      tei_mapping:
        element: "roleName"
        attributes:
          type: "honorific|official|occupation|civil|military|religious"
          nymRef: "URI reference to controlled vocabulary"
        related_elements:
          - "occupation"
          - "affiliation"
      nerd_class: null
      nerd_note: |
        NERD has no dedicated class for roles/occupations. Most NER systems
        conflate roles with persons. This convention separates them for
        precision in heritage contexts (staff roles, historical titles).        
    
    subcategories:
      # ----- OCCUPATIONAL ROLES -----
      OCCUPATION:
        code: "ROL.OCC"
        definition: "Professional occupations and job titles"
        examples:
          - "curator"
          - "archivist"
          - "librarian"
          - "conservator"
          - "director"
          - "registrar"
        ontology_class: "schema:Occupation"
        alternative_classes:
          - "tei:occupation"
        linkml_mapping:
          class_uri: "schema:Occupation"
          related_mappings:
            - "sdo:occupationalCategory"
        isco_note: |
          Link to ISCO-08 (International Standard Classification of Occupations)
          for standardized occupation codes when possible.          
        note: |
          Occupations are general categories of work.
          For specific positions in organizations, use ROL.POS.
          For the person holding the occupation, use AGENT.STF.          
      
      POSITION:
        code: "ROL.POS"
        definition: |
          Specific positions within organizations, typically unique roles
          that persist beyond the current holder.          
        examples:
          - "Director of the Rijksmuseum"
          - "Chief Curator of Prints"
          - "Head of Conservation"
          - "President of ICOM"
        ontology_class: "org:Post"
        org_properties:
          postIn: "org:postIn (links to organization)"
          heldBy: "org:heldBy (links to agent)"
        note: |
          Positions (Posts) are distinct from the people who hold them.
          org:Post models positions that exist independently of holders.
          Use when the position itself is referenced, not just the occupation.          
      
      # ----- HONORIFIC TITLES -----
      HONORIFIC:
        code: "ROL.HON"
        definition: |
          Honorific titles, forms of address, and ceremonial designations.          
        examples:
          - "Your Majesty"
          - "Professor"
          - "Dr."
          - "Sir"
          - "Dame"
          - "His Excellency"
          - "Reverend"
        ontology_class: "crm:E55_Type"
        tei_mapping:
          element: "roleName"
          attribute_type: "honorific"
        pnv_mapping: "pnv:honorificPrefix | pnv:honorificSuffix"
        note: |
          Honorifics often appear with names but are roles, not name parts:
          - "Professor Einstein" → AGENT.PER with ROL.HON
          - When analyzing name structure: APP.PNM with honorific component
          
          Academic degrees (PhD, MA) after names are honorificSuffix.          
      
      # ----- NOBILITY AND RANK -----
      NOBILITY:
        code: "ROL.NOB"
        definition: "Noble titles, aristocratic ranks, and royal designations"
        examples:
          - "Duke of Wellington"
          - "Countess of Blessington"
          - "Prince of Orange"
          - "Baron van Pallandt"
          - "King"
          - "Queen"
        ontology_class: "crm:E55_Type"
        tei_mapping:
          element: "roleName"
          attribute_type: "nobility"
        note: |
          Nobility titles may be:
          - Hereditary: Pass through family lineage
          - Created: Bestowed as honor
          - Extinct: Historical titles no longer in use
          
          Link to authority files for genealogical precision.          
      
      # ----- RELIGIOUS TITLES -----
      RELIGIOUS:
        code: "ROL.REL"
        definition: "Religious titles, orders, and ecclesiastical ranks"
        examples:
          - "Pope"
          - "Cardinal"
          - "Rabbi"
          - "Imam"
          - "Reverend"
          - "Brother"
          - "Sister"
          - "Abbot"
        ontology_class: "crm:E55_Type"
        tei_mapping:
          element: "roleName"
          attribute_type: "religious"
        note: |
          Religious roles span traditions:
          - Christian: Pope, Bishop, Deacon, Abbot, Prior
          - Jewish: Rabbi, Cantor
          - Islamic: Imam, Mufti, Ayatollah
          - Buddhist: Monk, Lama, Rinpoche          
      
      # ----- RELATIONAL ROLES -----
      RELATIONAL:
        code: "ROL.RLT"
        definition: |
          Relational roles defined by relationship to another agent:
          family relationships, apprenticeship, mentorship.          
        examples:
          - "father of Rembrandt"
          - "student of Frans Hals"
          - "wife of"
          - "heir to"
          - "successor of"
        ontology_class: "crm:E55_Type"
        bio_relations:
          family: "bio:parent, bio:child, bio:spouse, bio:sibling"
          professional: "bio:student, bio:mentor"
        note: |
          Relational roles require a reference agent:
          - "X's father" → X has role "father" relative to someone
          
          For family relationships, use BIO ontology properties.
          For professional relationships, use org:memberOf context.          
    
    inclusion_rules:
      - id: "ROL_INC001"
        rule: "Tag occupation terms when they identify professional function"
        examples:
          - "the curator stated..."
          - "as archivist, she organized..."
      
      - id: "ROL_INC002"
        rule: "Tag honorifics that precede or follow names"
        examples:
          - "Professor Einstein"
          - "Dr. Marie Curie"
          - "Jan de Wit, Director"
      
      - id: "ROL_INC003"
        rule: "Tag specific organizational positions"
        examples:
          - "Director of the Rijksmuseum"
          - "Chief Curator of Medieval Art"
      
      - id: "ROL_INC004"
        rule: "Tag nobility and religious titles"
        examples:
          - "Duke of Wellington"
          - "Cardinal Richelieu"
          - "Rabbi Akiva"
    
    exclusion_rules:
      - id: "ROL_EXC001"
        rule: "Do NOT tag generic person references as roles"
        examples:
          - "the man (not a role)"
          - "a woman (not a role)"
          - "someone (not a role)"
      
      - id: "ROL_EXC002"
        rule: "Do NOT tag organizations as roles"
        examples:
          - "the museum (use GROUP)"
          - "the university (use GROUP)"
      
      - id: "ROL_EXC003"
        rule: "Do NOT double-tag: person with role context → AGENT + ROLE"
        examples:
          - "'Director Jan de Wit' → AGENT.PER + ROL.POS"
        note: "Create separate annotations with linking"

  # ---------------------------------------------------------------------------
  # QUANTITY - Numeric and measurable values
  # ---------------------------------------------------------------------------
  QUANTITY:
    code: "QTY"
    definition: |
      Numerical quantities, measurements, counts, and ranges. Includes
      dimensions, populations, statistics, and any quantifiable values
      with or without units.      
    
    ontology_mappings:
      primary_class: "crm:E54_Dimension"
      alternative_classes:
        - "schema:QuantitativeValue"
        - "ivoa:Measure"
      nerd_class: "nerd:Amount"
      nerd_note: "NERD Amount class for all quantities and measurements."
    
    subcategories:
      COUNT:
        code: "QTY.CNT"
        definition: "Cardinal numbers and item counts"
        examples:
          - "8,000 artworks"
          - "over 1 million visitors"
          - "15 staff members"
        ontology_class: "xsd:integer"
      
      MEASUREMENT:
        code: "QTY.MSR"
        definition: "Physical measurements with units"
        examples:
          - "363 cm by 437 cm"
          - "2.5 meters tall"
          - "500 square meters"
        ontology_class: "crm:E54_Dimension"
      
      CURRENCY:
        code: "QTY.CUR"
        definition: "Monetary amounts"
        examples:
          - "15 euro admission"
          - "5 million euros"
          - "1.2 billion dollars"
        ontology_class: "schema:MonetaryAmount"
      
      PERCENTAGE:
        code: "QTY.PCT"
        definition: "Percentages and ratios"
        examples:
          - "75 percent of the collection"
          - "one-third of visitors"
        ontology_class: "xsd:decimal"
      
      RANGE:
        code: "QTY.RNG"
        definition: "Numeric ranges"
        examples:
          - "between 50,000 and 100,000"
          - "10-15 meters"
          - "ages 6-12"
        ontology_class: "schema:QuantitativeValue"
      
      ORDINAL:
        code: "QTY.ORD"
        definition: "Ordinal numbers and rankings"
        examples:
          - "the 5th largest museum"
          - "first place"
          - "17th century"
        ontology_class: "xsd:integer"
    
    inclusion_rules:
      - id: "QTY_INC001"
        rule: "Tag quantities with their units as single entities"
        examples:
          - "8,000 paintings (single entity)"
          - "2.5 million visitors annually"
      
      - id: "QTY_INC002"
        rule: "Tag approximate quantities"
        examples:
          - "approximately 500"
          - "over 1 million"
          - "nearly 10,000"
      
      - id: "QTY_INC003"
        rule: "Tag ranges as single entities"
        examples:
          - "10,000 to 15,000 items"
    
    exclusion_rules:
      - id: "QTY_EXC001"
        rule: "Do NOT tag numbers that are part of proper names"
        examples:
          - "World War 2 (part of name, tag as THING.EVT)"
          - "Henry VIII (part of name, tag as BEING)"
      
      - id: "QTY_EXC002"
        rule: "Do NOT tag page numbers, reference numbers, or identifiers"
        examples:
          - "page 42"
          - "inventory number 12345"

  # ---------------------------------------------------------------------------
  # TEMPORAL - Time references following TimeML/TIMEX3 model
  # ---------------------------------------------------------------------------
  # BREAKING CHANGE v1.7.0: Restructured from TEMPORAL_REFERENCE (TMP)
  # Now follows TimeML/TIMEX3 typology:
  #   - DATE: Calendar dates (absolute)
  #   - TIME: Clock times
  #   - DURATION: Temporal spans
  #   - SET: Recurring/periodic times
  # Plus absolute vs. relative distinction (TIMEX3 @mod attribute)
  # ---------------------------------------------------------------------------
  TEMPORAL:
    code: "TMP"
    definition: |
      Temporal expressions following the TimeML/TIMEX3 standard. Temporal
      expressions denote points, intervals, or frequencies on the timeline.
      
      Key distinctions:
      - ABSOLUTE: Resolved to specific calendar/clock ("15 July 1606")
      - RELATIVE: Requires context to resolve ("last year", "recently")
      - DURATION: Temporal extent ("three years", "for a decade")
      - SET: Recurring frequencies ("every Monday", "annually")
      
      TimeML/TIMEX3 is the ISO standard for temporal expression markup,
      widely used in NLP (TimeBank, TempEval). This provides interoperability
      with temporal reasoning systems.      
    
    design_rationale: |
      NERD's generic "Time" class lacks the semantic precision needed for:
      - Temporal reasoning and event ordering
      - Calendar normalization (Julian vs. Gregorian)
      - Duration calculations
      - Frequency detection (for opening hours, events)
      
      TimeML/TIMEX3 (ISO-TimeML) provides:
      - TYPE attribute: DATE | TIME | DURATION | SET
      - VALUE attribute: ISO 8601 normalized value
      - MOD attribute: START | END | MID | BEFORE | AFTER | APPROX
      
      CIDOC-CRM provides complementary modeling:
      - E52 Time-Span: Temporal extent with fuzzy boundaries
      - E61 Time Primitive: Precise instant or interval
      - E4 Period: Named historical periods      
    
    ontology_mappings:
      primary_class: "crm:E52_Time-Span"
      primary_class_definition: |
        CIDOC-CRM E52 Time-Span: "This class comprises abstract temporal extents,
        in the sense of Galilean physics, having a beginning, an end and a
        duration."        
      alternative_classes:
        - "time:TemporalEntity"  # W3C Time Ontology
        - "time:Instant"  # Specific point
        - "time:Interval"  # Duration with bounds
      linkml_mapping:
        class_uri: "crm:E52_Time-Span"
        exact_mappings:
          - "time:TemporalEntity"
        close_mappings:
          - "dct:temporal"
      nerd_class: "nerd:Time"
      nerd_deprecation_note: |
        DEPRECATED: NERD's Time class is too generic, lacking the DATE/TIME/
        DURATION/SET typology essential for temporal reasoning. Use TimeML
        TIMEX3 types for NLP annotation; map to CIDOC-CRM for semantics.
        Retain NERD mapping ONLY for basic NLP pipeline interchange.        
      timeml_mapping:
        element: "TIMEX3"
        attributes:
          type: "DATE | TIME | DURATION | SET"
          value: "ISO 8601 normalized value"
          mod: "BEFORE | AFTER | APPROX | START | END | MID"
          anchorTimeID: "ID of anchoring time for relative expressions"
      note: |
        W3C Time Ontology (OWL-Time) provides formal semantics for temporal
        entities compatible with CIDOC-CRM. TimeML provides NLP annotation
        conventions. Both map to ISO 8601 for interchange.        
    
    subcategories:
      # ----- ABSOLUTE TEMPORAL EXPRESSIONS -----
      DATE_ABS:
        code: "TMP.DAB"
        definition: |
          Absolute calendar dates that can be resolved without context.
          Maps to TimeML TIMEX3 type="DATE" without relative modifiers.          
        examples:
          - "15 July 1606"
          - "March 2023"
          - "1888"
          - "12/04/1943"
          - "the year 2000"
        ontology_class: "time:Instant"
        timeml_type: "DATE"
        xsd_type: "xsd:date | xsd:gYear | xsd:gYearMonth"
        linkml_mapping:
          class_uri: "time:Instant"
          exact_mappings:
            - "xsd:date"
            - "xsd:gYear"
        note: |
          Normalize to ISO 8601 format when possible:
          - Full date: 1606-07-15
          - Year-month: 2023-03
          - Year only: 1888
          
          Historical dates may require calendar specification (Julian/Gregorian).          
      
      DATE_REL:
        code: "TMP.DRL"
        definition: |
          Relative dates requiring context (document date, speech time)
          to resolve. Maps to TimeML TIMEX3 type="DATE" with anchorTimeID.          
        examples:
          - "last year"
          - "yesterday"
          - "next month"
          - "three years ago"
          - "in recent decades"
          - "since the war"
        ontology_class: "crm:E52_Time-Span"
        timeml_type: "DATE"
        timeml_anchor: "Requires @anchorTimeID to resolve"
        note: |
          Relative dates are common in heritage texts but require context:
          - "last year" → need document_date to resolve
          - "since the war" → need historical context (which war?)
          
          Mark with @certainty when resolution is uncertain.          
      
      TIME_ABS:
        code: "TMP.TAB"
        definition: |
          Absolute clock times. Maps to TimeML TIMEX3 type="TIME".          
        examples:
          - "10:00"
          - "14:30"
          - "midnight"
          - "noon"
          - "3 PM"
        ontology_class: "time:Instant"
        timeml_type: "TIME"
        xsd_type: "xsd:time"
        note: |
          Normalize to 24-hour ISO 8601 format: HH:MM:SS
          Named times: midnight → 00:00, noon → 12:00          
      
      TIME_REL:
        code: "TMP.TRL"
        definition: |
          Relative clock times requiring context.          
        examples:
          - "an hour later"
          - "that morning"
          - "in the evening"
        ontology_class: "crm:E52_Time-Span"
        timeml_type: "TIME"
      
      # ----- DURATIONS -----
      DURATION:
        code: "TMP.DUR"
        definition: |
          Temporal durations: lengths of time without fixed endpoints.
          Maps to TimeML TIMEX3 type="DURATION".          
        examples:
          - "three years"
          - "for a decade"
          - "two hours"
          - "the 17th century (100 years)"
          - "a fortnight"
          - "several months"
        ontology_class: "time:Duration"
        timeml_type: "DURATION"
        xsd_type: "xsd:duration"
        linkml_mapping:
          class_uri: "time:Duration"
          exact_mappings:
            - "xsd:duration"
        note: |
          Normalize to ISO 8601 duration format: P[n]Y[n]M[n]DT[n]H[n]M[n]S
          Examples:
          - "three years" → P3Y
          - "two hours" → PT2H
          - "a decade" → P10Y
          - "17th century" → P100Y (with temporal bounds for period)          
      
      # ----- RECURRING/PERIODIC TIMES -----
      SET:
        code: "TMP.SET"
        definition: |
          Recurring or periodic temporal expressions. Maps to TimeML
          TIMEX3 type="SET". Common for opening hours, event schedules.          
        examples:
          - "every Monday"
          - "annually"
          - "twice a week"
          - "open Tuesday-Sunday"
          - "each summer"
          - "quarterly"
        ontology_class: "schema:OpeningHoursSpecification"
        timeml_type: "SET"
        alternative_classes:
          - "time:GeneralDateTimeDescription"
        note: |
          SET expressions describe recurring patterns:
          - Frequency: "twice a week" → SET with @quant="2" @freq="1W"
          - Schedule: "every Monday" → SET with value="XXXX-WXX-1"
          
          For opening hours, schema:OpeningHoursSpecification provides
          structured representation with dayOfWeek, opens, closes properties.          
      
      # ----- OPENING HOURS (specialized SET) -----
      OPENHRS:
        code: "TMP.OPH"
        definition: |
          Institutional opening hours and operational schedules. A specialized
          form of SET expression.          
        examples:
          - "open Tuesday-Sunday 10:00-17:00"
          - "closed Mondays"
          - "last entry at 16:30"
          - "open daily except holidays"
        ontology_class: "schema:OpeningHoursSpecification"
        schema_properties:
          dayOfWeek: "schema:DayOfWeek (Monday, Tuesday, etc.)"
          opens: "xsd:time (opening time)"
          closes: "xsd:time (closing time)"
          validFrom: "xsd:date (seasonal start)"
          validThrough: "xsd:date (seasonal end)"
        note: |
          Links to GROUP hypernym (heritage institutions) via schema:openingHours.
          Use schema:specialOpeningHoursSpecification for holidays/exceptions.          
      
      # ----- DATE RANGES -----
      RANGE:
        code: "TMP.RNG"
        definition: |
          Date/time ranges with explicit start and end points.          
        examples:
          - "1888-1890"
          - "from March to June 2023"
          - "10 February - 4 June 2023"
          - "between 1650 and 1670"
        ontology_class: "time:ProperInterval"
        alternative_classes:
          - "crm:E52_Time-Span"
        edtf_note: |
          Extended Date/Time Format (EDTF, ISO 8601-2) provides:
          - Intervals: 1888/1890
          - Open ranges: 1888/.. (from 1888 onwards)
          - Uncertain: 1888?/1890          
      
      # ----- NAMED PERIODS -----
      CENTURY:
        code: "TMP.CEN"
        definition: "Century references, a common periodization"
        examples:
          - "17th century"
          - "the 1800s"
          - "nineteenth century"
          - "early 20th century"
        ontology_class: "crm:E52_Time-Span"
        timeml_type: "DATE"
        note: |
          Normalize centuries to date ranges:
          - "17th century" → 1601-01-01/1700-12-31
          - "the 1800s" → 1800-01-01/1899-12-31
          
          Modifiers (early, mid, late) narrow the range.          
      
      ERA:
        code: "TMP.ERA"
        definition: |
          Named historical periods, movements, and eras. These are cultural
          periodizations, not calendar units.          
        examples:
          - "the Golden Age"
          - "the Renaissance"
          - "Medieval period"
          - "Edo period"
          - "the Enlightenment"
          - "Art Deco era"
        ontology_class: "crm:E4_Period"
        linkml_mapping:
          class_uri: "crm:E4_Period"
          close_mappings:
            - "dct:PeriodOfTime"
        note: |
          Named periods have fuzzy boundaries and geographic variation:
          - "Renaissance" varies by region (Italy vs. Northern Europe)
          - "Golden Age" is culture-specific (Dutch vs. Spanish)
          
          Use crm:E4_Period which explicitly allows fuzzy temporal boundaries.
          Link to authority files (Getty AAT, Wikidata) for disambiguation.          
      
      EXHIBPER:
        code: "TMP.EXP"
        definition: "Exhibition periods and event dates"
        examples:
          - "10 February - 4 June 2023"
          - "on view through December"
          - "opening reception: May 5, 7-9 PM"
        ontology_class: "crm:E52_Time-Span"
        schema_mapping: "schema:Event with schema:startDate and schema:endDate"
        note: "Use for temporally bounded institutional events."
    
    inclusion_rules:
      - id: "TMP_INC001"
        rule: "Tag complete date/time expressions as single entities"
        examples:
          - "15 July 1606 (single entity)"
          - "between 1888 and 1890 (range)"
          - "every Monday at 10:00"
      
      - id: "TMP_INC002"
        rule: "Tag named periods and eras"
        examples:
          - "the Dutch Golden Age"
          - "the Baroque period"
          - "during the Renaissance"
      
      - id: "TMP_INC003"
        rule: "Tag opening hours as complete SET expressions"
        examples:
          - "Tuesday to Sunday, 10:00-17:00"
          - "open daily except Mondays"
      
      - id: "TMP_INC004"
        rule: "Tag relative expressions with their anchor context"
        examples:
          - "last year (relative to document date)"
          - "since the merger (relative to event)"
      
      - id: "TMP_INC005"
        rule: "Tag durations even without specific anchoring"
        examples:
          - "for three years"
          - "a decade of research"
    
    exclusion_rules:
      - id: "TMP_EXC001"
        rule: "Do NOT tag deictics without recoverable reference"
        examples:
          - "now (unanchored)"
          - "today (unless document date known)"
        note: "These require pragmatic resolution beyond text"
      
      - id: "TMP_EXC002"
        rule: "Do NOT tag ordinal centuries as QUANTITY"
        examples:
          - "17th century → TMP.CEN, not QTY.ORD"
      
      - id: "TMP_EXC003"
        rule: "Do NOT tag temporal prepositions alone"
        examples:
          - "before (preposition, not temporal reference)"
          - "during (connector)"

  # ---------------------------------------------------------------------------
  # WORK - Intellectual/creative works following FRBR/LRM/FRBRoo model
  # ---------------------------------------------------------------------------
  # BREAKING CHANGE v1.7.0: Renamed from TEXTUAL_REFERENCE (TXT) to WORK (WRK)
  # and restructured following FRBR/IFLA-LRM model:
  #   - WORK: Abstract intellectual creation (the concept)
  #   - EXPRESSION: Realization in a specific form (the text)
  #   - MANIFESTATION: Physical embodiment (the publication)
  #   - ITEM: Single exemplar (the specific copy)
  # 
  # NERD's "Product" class is semantically wrong for intellectual works.
  # FRBR provides the bibliographic and archival standard.
  # ---------------------------------------------------------------------------
  WORK:
    code: "WRK"
    definition: |
      References to intellectual and creative works, spanning the FRBR/LRM
      abstraction levels from abstract concept (Work) to physical copy (Item).
      
      The FRBR model distinguishes:
      - WORK: Distinct intellectual creation ("Hamlet" as a concept)
      - EXPRESSION: Specific intellectual realization ("Hamlet in Dutch translation")
      - MANIFESTATION: Physical embodiment ("2023 Penguin edition of Hamlet")
      - ITEM: Single exemplar ("the copy in Amsterdam Public Library")
      
      Most textual mentions reference WORK level (titles, concepts).
      Archival contexts often reference MANIFESTATION or ITEM level.      
    
    design_rationale: |
      NERD's "Product" class conflates:
      - Creative works (books, music, films)
      - Commercial products (software, merchandise)
      - Services (subscriptions, access)
      
      For heritage contexts, FRBR (now IFLA-LRM) is the bibliographic standard:
      - Library catalogs use FRBR for bibliographic relationships
      - FRBRoo integrates with CIDOC-CRM for museum contexts
      - RiC-O (archival) aligns with FRBR for records
      
      This enables:
      - Linking editions, translations, and copies of the same work
      - Tracking provenance of specific items
      - Distinguishing conceptual references from physical objects      
    
    ontology_mappings:
      primary_class: "frbroo:F1_Work"
      primary_class_definition: |
        FRBRoo F1 Work: "This class comprises distinct concepts or combination
        of concepts identified in artistic and intellectual expressions."
        Equivalent to IFLA-LRM Work.        
      alternative_classes:
        - "lrm:Work"  # IFLA Library Reference Model
        - "crm:E89_Propositional_Object"  # CIDOC-CRM intellectual content
        - "schema:CreativeWork"  # Schema.org
        - "dct:BibliographicResource"  # Dublin Core
      linkml_mapping:
        class_uri: "frbroo:F1_Work"
        exact_mappings:
          - "lrm:Work"
        close_mappings:
          - "schema:CreativeWork"
          - "crm:E89_Propositional_Object"
        related_mappings:
          - "bf:Work"  # BIBFRAME
      nerd_class: "nerd:Product"
      nerd_deprecation_note: |
        DEPRECATED: NERD's Product class is semantically incorrect for
        intellectual works. "Product" implies commercial goods, not creative
        expressions. Use FRBR/FRBRoo/LRM for bibliographic precision.
        Retain NERD mapping ONLY for basic NLP pipeline interchange.        
      frbr_mapping:
        work: "frbroo:F1_Work | lrm:Work"
        expression: "frbroo:F2_Expression | lrm:Expression"
        manifestation: "frbroo:F3_Manifestation | lrm:Manifestation"
        item: "frbroo:F5_Item | lrm:Item"
      note: |
        FRBRoo is the object-oriented version of FRBR, aligned with CIDOC-CRM.
        IFLA-LRM (2017) is the consolidated successor to FRBR/FRAD/FRSAD.
        BIBFRAME (bf:) is the Library of Congress RDF vocabulary for bibliographic data.        
    
    subcategories:
      # ----- FRBR WORK LEVEL (abstract concept) -----
      WORK_ABSTRACT:
        code: "WRK.ABS"
        definition: |
          Abstract intellectual works referenced by title or concept,
          independent of specific editions or copies.          
        examples:
          - "Hamlet"
          - "The Diary of Anne Frank"
          - "Beethoven's Ninth Symphony"
          - "the Quran"
        ontology_class: "frbroo:F1_Work"
        alternative_classes:
          - "lrm:Work"
          - "bf:Work"
        linkml_mapping:
          class_uri: "frbroo:F1_Work"
          exact_mappings:
            - "lrm:Work"
            - "bf:Work"
        note: |
          Use for references to works as concepts, not specific editions:
          - "Shakespeare's Hamlet" (the work, not a specific edition)
          - "Anne Frank's diary" (the conceptual work)
          
          Link to authority files (VIAF, Wikidata) for work identification.          
      
      # ----- FRBR EXPRESSION LEVEL (specific realization) -----
      EXPRESSION:
        code: "WRK.EXP"
        definition: |
          Specific realizations of works: translations, versions, arrangements,
          performances captured as recordings.          
        examples:
          - "the Dutch translation of Hamlet"
          - "the 1603 quarto text of Hamlet"
          - "Karajan's 1962 recording of Beethoven's Ninth"
          - "the King James Bible"
        ontology_class: "frbroo:F2_Expression"
        alternative_classes:
          - "lrm:Expression"
          - "bf:Work"  # BIBFRAME conflates Work/Expression
        note: |
          Expressions are realizations of Works:
          - Same Work, different language → different Expression
          - Same Work, different arrangement → different Expression
          - Same musical Work, different performance → different Expression          
      
      # ----- FRBR MANIFESTATION LEVEL (publication/edition) -----
      MANIFESTATION:
        code: "WRK.MAN"
        definition: |
          Physical or digital embodiments: editions, publications, pressings,
          broadcasts. Manifestations are producible in multiple copies.          
        examples:
          - "the 2023 Penguin edition of Hamlet"
          - "Museum Journal vol. 15, no. 2"
          - "Nature magazine, March 2023"
          - "the 1609 Quarto of Shakespeare's Sonnets"
        ontology_class: "frbroo:F3_Manifestation"
        alternative_classes:
          - "lrm:Manifestation"
          - "bf:Instance"  # BIBFRAME equivalent
          - "schema:Book"
          - "schema:Periodical"
        linkml_mapping:
          class_uri: "frbroo:F3_Manifestation"
          exact_mappings:
            - "lrm:Manifestation"
            - "bf:Instance"
          close_mappings:
            - "schema:Book"
        note: |
          Manifestations are what libraries catalog and acquire:
          - Publisher, date, ISBN, format are Manifestation attributes
          - Multiple identical copies = one Manifestation, many Items          
      
      # ----- FRBR ITEM LEVEL (specific copy) -----
      ITEM:
        code: "WRK.ITM"
        definition: |
          Single exemplars: a specific copy, with provenance, annotations,
          damage. Items are unique physical objects.          
        examples:
          - "the British Library's copy of the First Folio"
          - "manuscript KB 128 A 14 in the Royal Library"
          - "the annotated copy formerly owned by Van Gogh"
        ontology_class: "frbroo:F5_Item"
        alternative_classes:
          - "lrm:Item"
          - "bf:Item"
          - "crm:E22_Human-Made_Object"
        note: |
          Items have individual provenance and condition:
          - Ownership history, annotations, damage
          - Physical location (shelf mark, repository)
          - For manuscripts, each copy is a unique Item          
      
      # ----- MANUSCRIPT (always Item level) -----
      MANUSCRIPT:
        code: "WRK.MSS"
        definition: |
          Handwritten documents and manuscripts. Manuscripts are unique
          Items (FRBR Item level), not reproducible Manifestations.          
        examples:
          - "Codex Manesse"
          - "the Book of Kells"
          - "Leonardo's notebooks"
          - "the Dead Sea Scrolls"
        ontology_class: "crm:E22_Human-Made_Object"
        alternative_classes:
          - "frbroo:F5_Item"
          - "rico:Record"
        note: |
          Manuscripts are physical objects that embody intellectual content:
          - Use crm:E22_Human-Made_Object for the physical codex/scroll
          - Use frbroo:F1_Work for the textual content
          - Link via frbroo:R4_comprises_carriers_of          
      
      # ----- ARCHIVAL RECORDS -----
      ARCHIVAL:
        code: "WRK.ARC"
        definition: |
          Archival records, documents, and files. May be single items or
          aggregations (fonds, series).          
        examples:
          - "deed of sale dated 1650"
          - "birth certificate of Rembrandt"
          - "notarial act NL-SAA 5075/2135"
          - "the VOC archives"
        ontology_class: "rico:Record"
        alternative_classes:
          - "rico:RecordSet"  # For aggregations (fonds, series)
          - "crm:E31_Document"
        rico_levels:
          fonds: "rico:RecordSet with rico:hasRecordSetType 'fonds'"
          series: "rico:RecordSet with rico:hasRecordSetType 'series'"
          file: "rico:RecordSet with rico:hasRecordSetType 'file'"
          item: "rico:Record"
        note: |
          RiC-O provides archival hierarchy:
          - Fonds → Series → File → Item
          - Use rico:isOrWasIncludedIn for containment
          - Use rico:hasOrHadIdentifier for call numbers          
      
      # ----- DIGITAL RESOURCES -----
      WEBSITE:
        code: "WRK.WEB"
        definition: "Web pages and online resources as intellectual works"
        examples:
          - "the museum's online collection"
          - "the Wikipedia article on Rembrandt"
          - "the Rijksmuseum website"
        ontology_class: "schema:WebPage"
        alternative_classes:
          - "schema:WebSite"
          - "fabio:WebPage"  # FRBR-aligned Bibliographic Ontology
        note: |
          Websites are Manifestations of digital Works.
          Archived versions (Internet Archive) are different Manifestations
          of the same Expression.          
      
      URL:
        code: "WRK.URL"
        definition: "URLs as identifiers for digital resources"
        examples:
          - "www.rijksmuseum.nl"
          - "https://www.britishmuseum.org"
          - "https://doi.org/10.1000/xyz123"
        ontology_class: "schema:URL"
        note: |
          URLs are locators, not works themselves.
          Links to GROUP hypernym via schema:url.
          DOIs are persistent identifiers for Works/Expressions.          
      
      # ----- CONTACT INFORMATION (not works, but often co-occur) -----
      EMAIL:
        code: "WRK.EML"
        definition: "Email addresses as contact identifiers"
        examples:
          - "info@rijksmuseum.nl"
          - "contact@museum.org"
        ontology_class: "schema:email"
        note: |
          Email addresses are identifiers, not intellectual works.
          Included here for practical extraction convenience.
          Links to GROUP hypernym via schema:email.          
      
      SOCIAL:
        code: "WRK.SOC"
        definition: "Social media handles and profiles"
        examples:
          - "@rijksmuseum"
          - "facebook.com/britishmuseum"
          - "Instagram: @vangoghmuseum"
        ontology_class: "foaf:OnlineAccount"
        foaf_mapping:
          class: "foaf:OnlineAccount"
          properties:
            account_name: "foaf:accountName"
            service_homepage: "foaf:accountServiceHomepage"
        note: |
          Social accounts are identifiers for agent presence on platforms.
          Links to GROUP hypernym via foaf:account.          
      
      # ----- CITATIONS -----
      CITATION:
        code: "WRK.CIT"
        definition: "Bibliographic citations and references"
        examples:
          - "Smith (2020)"
          - "cf. Van Gogh Letters, no. 532"
          - "[ibid., p. 42]"
        ontology_class: "dct:bibliographicCitation"
        alternative_classes:
          - "cito:Citation"  # Citation Typing Ontology
        note: |
          Citations are references to Works/Expressions/Manifestations.
          CiTO (Citation Typing Ontology) provides citation intent types:
          - cito:cites, cito:critiques, cito:supports, etc.          
    
    inclusion_rules:
      - id: "WRK_INC001"
        rule: "Tag work titles at appropriate FRBR level"
        examples:
          - "'Hamlet' (Work level - the concept)"
          - "'2023 Penguin edition' (Manifestation level)"
          - "'the British Library's copy' (Item level)"
      
      - id: "WRK_INC002"
        rule: "Tag URLs and identifiers as complete strings"
        examples:
          - "info@rijksmuseum.nl"
          - "https://www.rijksmuseum.nl"
          - "doi:10.1000/xyz123"
      
      - id: "WRK_INC003"
        rule: "Tag archival references with call numbers when present"
        examples:
          - "NL-SAA 5075/2135"
          - "KB 128 A 14"
      
      - id: "WRK_INC004"
        rule: "Tag manuscripts as Items (unique objects)"
        examples:
          - "Codex Manesse"
          - "the Book of Kells"
    
    exclusion_rules:
      - id: "WRK_EXC001"
        rule: "Do NOT tag generic work type references"
        examples:
          - "the document (no title)"
          - "a manuscript (not named)"
          - "some book"
      
      - id: "WRK_EXC002"
        rule: "Do NOT tag UI navigation elements"
        examples:
          - "Click here"
          - "Read more"
          - "Download PDF"
      
      - id: "WRK_EXC003"
        rule: "Do NOT confuse physical carrier with intellectual work"
        examples:
          - "the book (physical object → use THING if just the object)"
          - "the paper (material, not work)"

  # ---------------------------------------------------------------------------
  # THING - Objects, concepts, and phenomena
  # ---------------------------------------------------------------------------
  THING:
    code: "THG"
    definition: |
      Named objects, artifacts, concepts, events, and other entities that
      do not fit the above categories. Broad category for tangible items
      and abstract concepts.      
    
    ontology_mappings:
      primary_class: "crm:E70_Thing"
      alternative_classes:
        - "schema:Thing"
        - "edm:NonInformation_Resource"
      nerd_class: "nerd:Thing"
      nerd_subclasses:
        - "nerd:Product"
        - "nerd:Event"
      nerd_note: |
        NERD Thing is the base class. Use nerd:Product for objects/artifacts,
        nerd:Event for THG.EVT subcategory.        
      edm_note: |
        Europeana Data Model (EDM) provides cultural heritage-specific classes:
        - edm:ProvidedCHO: Cultural Heritage Object (the real-world thing)
        - edm:PhysicalThing: Physical/material objects
        - edm:Event: Historical or cultural events
        - edm:TimeSpan: Temporal extent of events/objects        
    
    subcategories:
      ARTWORK:
        code: "THG.ART"
        definition: "Physical artworks and art objects"
        examples:
          - "The Night Watch painting"
          - "Rodin's The Thinker"
          - "the Rosetta Stone"
        ontology_class: "crm:E22_Man-Made_Object"
        alternative_classes:
          - "edm:PhysicalThing"
          - "edm:ProvidedCHO"
        edm_note: |
          For artworks in cultural heritage contexts:
          - edm:ProvidedCHO for the conceptual cultural object
          - edm:PhysicalThing for the material manifestation          
      
      ARTIFACT:
        code: "THG.AFT"
        definition: "Historical objects and artifacts"
        examples:
          - "the crown jewels"
          - "Viking helmet"
          - "Roman coin"
        ontology_class: "crm:E22_Man-Made_Object"
        alternative_classes:
          - "edm:PhysicalThing"
      
      SPECIES:
        code: "THG.SPC"
        definition: "Biological species and taxa"
        examples:
          - "Tyrannosaurus rex"
          - "Quercus robur"
          - "dodo"
        ontology_class: "schema:Taxon"
      
      EVENT:
        code: "THG.EVT"
        definition: "Named historical events"
        examples:
          - "World War II"
          - "the French Revolution"
          - "the Great Fire of London"
        ontology_class: "crm:E5_Event"
        alternative_classes:
          - "edm:Event"
          - "schema:Event"
        edm_properties:
          has_time_span: "edm:happenedAt"
          occurred_at_place: "edm:occurredAt"
        note: |
          EDM Event extends CIDOC-CRM E5_Event for cultural heritage contexts.
          Use edm:happenedAt for temporal scope, edm:occurredAt for location.          
      
      CONCEPT:
        code: "THG.CON"
        definition: "Abstract concepts, movements, styles"
        examples:
          - "Impressionism"
          - "the Baroque style"
          - "human rights"
        ontology_class: "crm:E55_Type"
        alternative_classes:
          - "skos:Concept"
      
      LANGUAGE:
        code: "THG.LNG"
        definition: "Named languages"
        examples:
          - "Dutch"
          - "Latin"
          - "Old French"
        ontology_class: "crm:E56_Language"
      
      MATERIAL:
        code: "THG.MAT"
        definition: "Materials and substances"
        examples:
          - "oil on canvas"
          - "bronze"
          - "parchment"
        ontology_class: "crm:E57_Material"
    
    inclusion_rules:
      - id: "THG_INC001"
        rule: "Tag named artworks as physical objects (distinct from their titles)"
        examples:
          - "the painting known as The Night Watch"
          - "Michelangelo's David"
      
      - id: "THG_INC002"
        rule: "Tag named historical events"
        examples:
          - "the Eighty Years War"
          - "the Golden Age"
      
      - id: "THG_INC003"
        rule: "Tag art movements and styles when named"
        examples:
          - "Dutch Golden Age painting"
          - "Art Nouveau"
    
    exclusion_rules:
      - id: "THG_EXC001"
        rule: "Do NOT tag generic object references"
        examples:
          - "the object (not named)"
          - "an artifact (not specific)"
      
      - id: "THG_EXC002"
        rule: "Do NOT tag common nouns unless part of a proper name"
        examples:
          - "painting (generic)"
          - "sculpture (generic)"

# =============================================================================
# SECTION 3: UNIVERSAL EXCLUSION RULES
# =============================================================================
# These exclusion rules apply to ALL entity types across all text sources.

universal_exclusions:

  navigation:
    id: "UNI_EXC001"
    description: "Navigation elements and menu items"
    patterns:
      - "Home"
      - "About Us"
      - "Contact"
      - "Menu"
      - "Search"
      - "Login"
      - "Sign Up"
      - "Back to top"
      - "Next"
      - "Previous"
      - "Read more"
      - "Learn more"
      - "Click here"
      - "See all"
      - "View more"
      - "Show less"
    note: "These are UI elements, not named entities"

  calls_to_action:
    id: "UNI_EXC002"
    description: "Marketing and call-to-action phrases"
    patterns:
      - "Buy now"
      - "Subscribe"
      - "Book tickets"
      - "Plan your visit"
      - "Become a member"
      - "Donate"
      - "Shop"
      - "Get tickets"
      - "Reserve"
      - "Join us"
    note: "Action prompts, not entity references"

  social_boilerplate:
    id: "UNI_EXC003"
    description: "Generic social media text"
    patterns:
      - "Follow us"
      - "Share"
      - "Like"
      - "Tweet"
      - "Pin it"
      - "Share on Facebook"
      - "Follow on Instagram"
    note: "Tag actual handles (TXT.SOC), not these generic phrases"

  technical_artifacts:
    id: "UNI_EXC004"
    description: "Content management system artifacts"
    patterns:
      - "Posted by"
      - "Last updated"
      - "Tags:"
      - "Categories:"
      - "Comments"
      - "Leave a reply"
      - "Related posts"
      - "Powered by"
      - "Cookie settings"
      - "Privacy policy"
      - "Terms of use"
      - "All rights reserved"
    note: "Technical/legal boilerplate, not content entities"

  generic_words:
    id: "UNI_EXC005"
    description: "Common words that should not be tagged"
    patterns:
      - "Welcome"
      - "Information"
      - "Details"
      - "Overview"
      - "Introduction"
      - "Summary"
      - "Description"
      - "Features"
      - "Highlights"
      - "News"
      - "Events"
      - "Updates"
    note: "Section headers and generic labels, not named entities"

  pronouns:
    id: "UNI_EXC006"
    description: "Pronouns and demonstrative words"
    patterns:
      - "he"
      - "she"
      - "it"
      - "they"
      - "we"
      - "you"
      - "I"
      - "him"
      - "her"
      - "them"
      - "us"
      - "this"
      - "that"
      - "these"
      - "those"
      - "here"
      - "there"
      - "where"
    note: "Pronouns require coreference resolution, not NER"

# =============================================================================
# SECTION 4: DOUBLE-TAGGING PATTERNS
# =============================================================================
# Some text spans warrant multiple entity tags when they serve multiple
# semantic functions. Apply sparingly and document clearly.

double_tagging:
  
  description: |
    Double-tagging is permitted when a single text span genuinely serves
    multiple semantic functions. Each tag must be independently justified.
    Do NOT double-tag for redundancy.    
  
  permitted_patterns:
    
    - pattern: "BEING + DENOMINATION"
      description: "Person reference with structured name analysis"
      example:
        text: "Rembrandt Harmenszoon van Rijn"
        tags:
          - type: "BEING.PER"
            span: "Rembrandt Harmenszoon van Rijn"
            purpose: "Identifies the person"
          - type: "DEN.PNM"
            span: "Rembrandt Harmenszoon van Rijn"
            purpose: "Analyzes name structure"
            pnv_components:
              givenName: "Rembrandt"
              patronym: "Harmenszoon"
              surnamePrefix: "van"
              baseSurname: "Rijn"
    
    - pattern: "ORGANISATION + PLACE"
      description: "Institution that is also a place/building"
      example:
        text: "the Rijksmuseum"
        tags:
          - type: "ORG.HER"
            span: "the Rijksmuseum"
            purpose: "Identifies the organization"
          - type: "PLC.BLD"
            span: "the Rijksmuseum"
            purpose: "Identifies the physical building"
      note: "Apply when both organizational and locational aspects are relevant"
    
    - pattern: "THING.ART + DENOMINATION"
      description: "Artwork object with its title"
      example:
        text: "The Night Watch"
        tags:
          - type: "THG.ART"
            span: "The Night Watch"
            purpose: "The physical painting"
          - type: "DEN.TIT"
            span: "The Night Watch"
            purpose: "The title of the work"
    
    - pattern: "TEMPORAL + THING.EVT"
      description: "Event that is also a time reference"
      example:
        text: "World War II"
        tags:
          - type: "THG.EVT"
            span: "World War II"
            purpose: "The historical event"
          - type: "TMP.ERA"
            span: "World War II"
            purpose: "The time period (1939-1945)"

  prohibited_patterns:
    
    - pattern: "Same category double-tag"
      description: "Never apply two tags from the same category"
      example:
        text: "Amsterdam"
        wrong: "PLC.SET + PLC.CTY (Both are PLACE subcategories)"
        correct: "PLC.SET (Choose the most specific applicable type)"
    
    - pattern: "Redundant hierarchy"
      description: "Do not tag both a whole and its parts"
      example:
        text: "Rijksmuseum, Museumstraat 1, Amsterdam"
        wrong: "tag entire span as PLC.IAD AND also tag Amsterdam separately as PLC.SET"
        correct: "tag entire span as PLC.IAD (includes city)"
        note: "Subordinate elements are implicit in the containing entity"

# =============================================================================
# SECTION 5: ONTOLOGY RELATIONSHIP PATTERNS
# =============================================================================
# Bidirectional relationships between entity types, with ontology predicates.

relationships:

  description: |
    Relationships connect entities extracted from text. Each relationship
    has a subject, predicate, and object, with ontology mappings for RDF
    serialization. Relationships are bidirectional where applicable.    

  patterns:

    - name: "staff_member_of"
      subject_type: "BEING.STF"
      predicate: "org:memberOf"
      inverse_predicate: "org:hasMember"
      object_type: "ORG.HER"
      example:
        text: "Dr. Maria van den Berg, Director of the Rijksmuseum"
        subject: "Dr. Maria van den Berg"
        object: "Rijksmuseum"
    
    - name: "governed_by"
      subject_type: "ORG.HER"
      predicate: "org:subOrganizationOf"
      inverse_predicate: "org:hasSubOrganization"
      object_type: "ORG.PAR"
      example:
        text: "The Rijksmuseum, part of the Ministry of Culture"
        subject: "Rijksmuseum"
        object: "Ministry of Culture"
    
    - name: "located_in"
      subject_type: "ORG.HER"
      predicate: "schema:location"
      inverse_predicate: "schema:containsPlace"
      object_type: "PLC"
      example:
        text: "The British Museum in London"
        subject: "British Museum"
        object: "London"
    
    - name: "has_address"
      subject_type: "ORG.HER"
      predicate: "schema:address"
      object_type: "PLC.IAD"
      example:
        text: "Rijksmuseum, Museumstraat 1, 1071 XX Amsterdam"
        subject: "Rijksmuseum"
        object: "Museumstraat 1, 1071 XX Amsterdam"
    
    - name: "has_email"
      subject_type: "ORG.HER"
      predicate: "schema:email"
      object_type: "TXT.IEM"
      example:
        subject: "Rijksmuseum"
        object: "info@rijksmuseum.nl"
    
    - name: "has_website"
      subject_type: "ORG.HER"
      predicate: "schema:url"
      object_type: "TXT.IWB"
      example:
        subject: "Rijksmuseum"
        object: "https://www.rijksmuseum.nl"
    
    - name: "has_social_media"
      subject_type: "ORG.HER"
      predicate: "schema:sameAs"
      object_type: "TXT.SOC"
      example:
        subject: "Rijksmuseum"
        object: "@rijksmuseum"
    
    - name: "has_collection"
      subject_type: "ORG.HER"
      predicate: "rico:hasOrHadAllMembersOf"
      inverse_predicate: "rico:isOrWasIncludedIn"
      object_type: "DEN.COL"
      example:
        text: "The Rijksmuseum Golden Age Collection"
        subject: "Rijksmuseum"
        object: "Golden Age Collection"
    
    - name: "created_by"
      subject_type: "THG.ART"
      predicate: "crm:P14_carried_out_by"
      inverse_predicate: "crm:P14i_performed"
      object_type: "BEING.PER"
      example:
        text: "The Night Watch by Rembrandt"
        subject: "The Night Watch"
        object: "Rembrandt"
    
    - name: "has_title"
      subject_type: "THG.ART"
      predicate: "crm:P102_has_title"
      object_type: "DEN.TIT"
      example:
        subject: "the painting"
        object: "The Night Watch"
    
    - name: "occurred_during"
      subject_type: "THG.EVT"
      predicate: "crm:P4_has_time-span"
      object_type: "TMP"
      example:
        text: "World War II (1939-1945)"
        subject: "World War II"
        object: "1939-1945"
    
    - name: "exhibition_dates"
      subject_type: "DEN.EXH"
      predicate: "crm:P4_has_time-span"
      object_type: "TMP.EXP"
      example:
        text: "Vermeer exhibition, 10 February - 4 June 2023"
        subject: "Vermeer exhibition"
        object: "10 February - 4 June 2023"
    
    - name: "has_opening_hours"
      subject_type: "ORG.HER"
      predicate: "schema:openingHoursSpecification"
      object_type: "TMP.OPH"
      example:
        text: "The museum is open Tuesday-Sunday, 10:00-17:00"
        subject: "The museum"
        object: "Tuesday-Sunday, 10:00-17:00"
    
    # -------------------------------------------------------------------------
    # Additional W3C Org Ontology Relationship Patterns
    # -------------------------------------------------------------------------
    
    - name: "has_site"
      subject_type: "ORG.HER"
      predicate: "org:hasSite"
      inverse_predicate: "org:siteOf"
      object_type: "PLC.IAD"
      ontology: "org"
      example:
        text: "The Rijksmuseum has locations at Museumstraat and Schiphol Airport"
        subject: "Rijksmuseum"
        object: "Museumstraat location"
      note: |
        org:Site represents an office or premise at which the organization
        is located. Use for physical institutional locations. Distinct from
        schema:address which captures postal address data.        
    
    - name: "has_unit"
      subject_type: "ORG.HER"
      predicate: "org:hasUnit"
      inverse_predicate: "org:unitOf"
      object_type: "ORG.UNT"
      ontology: "org"
      example:
        text: "The Rijksmuseum's Department of Prints and Drawings"
        subject: "Rijksmuseum"
        object: "Department of Prints and Drawings"
      note: |
        Links organizational units to their parent organization.
        Use for named internal divisions that function as distinct entities.        
    
    - name: "head_of"
      subject_type: "BEING.STF"
      predicate: "org:headOf"
      object_type: "ORG.HER"
      ontology: "org"
      example:
        text: "Taco Dibbits, Director of the Rijksmuseum"
        subject: "Taco Dibbits"
        object: "Rijksmuseum"
      note: "Use for leadership roles. More specific than org:memberOf."
    
    - name: "holds_post"
      subject_type: "BEING.STF"
      predicate: "org:holds"
      inverse_predicate: "org:heldBy"
      object_type: "org:Post"
      ontology: "org"
      example:
        text: "Dr. van den Berg holds the position of Chief Curator"
        subject: "Dr. van den Berg"
        object: "Chief Curator"
      note: |
        org:Post represents a position that exists independently of the
        person filling it. Use when position titles are formally defined.        
    
    - name: "has_classification"
      subject_type: "ORG.HER"
      predicate: "org:classification"
      object_type: "THG.CON"
      ontology: "org"
      example:
        text: "The institution is classified as a national museum"
        subject: "The institution"
        object: "national museum"
      note: |
        org:classification links to a classification scheme category.
        Use for institution type classifications (museum, archive, library).        
    
    - name: "member_during"
      subject_type: "org:Membership"
      predicate: "org:memberDuring"
      object_type: "TMP"
      ontology: "org"
      example:
        text: "She served as curator from 2010 to 2015"
        subject: "membership"
        object: "2010-2015"
      note: |
        Temporal scope of a membership/role. Uses org:Membership as
        n-ary relationship between person, organization, and role.        
    
    - name: "collaborates_with"
      subject_type: "ORG.HER"
      predicate: "org:linkedTo"
      inverse_predicate: "org:linkedTo"
      object_type: "ORG.HER"
      ontology: "org"
      example:
        text: "The Rijksmuseum collaborates with the Van Gogh Museum"
        subject: "Rijksmuseum"
        object: "Van Gogh Museum"
      note: |
        org:linkedTo indicates a general relationship between organizations.
        Use for partnerships, collaborations, and informal associations.        
    
    - name: "resulted_from_change"
      subject_type: "ORG.HER"
      predicate: "org:resultedFrom"
      inverse_predicate: "org:resultingOrganization"
      object_type: "org:ChangeEvent"
      ontology: "org"
      example:
        text: "The Noord-Hollands Archief was formed through a merger in 2001"
        subject: "Noord-Hollands Archief"
        object: "2001 merger event"
      note: |
        Tracks organizational change events (mergers, splits, reorganizations).
        Use with org:ChangeEvent for provenance of institutional changes.        
    
    - name: "changed_by"
      subject_type: "ORG.HER"
      predicate: "org:changedBy"
      object_type: "org:ChangeEvent"
      ontology: "org"
      example:
        text: "The archive was restructured in 2018"
        subject: "The archive"
        object: "2018 restructuring"
      note: "Use for change events that affected but didn't create the organization."
    
    # -------------------------------------------------------------------------
    # RegOrg (Registered Organization) Relationship Patterns
    # -------------------------------------------------------------------------
    
    - name: "has_registration"
      subject_type: "ORG.HER"
      predicate: "rov:registration"
      object_type: "rov:Identifier"
      ontology: "rov"
      example:
        text: "Rijksmuseum (KvK 41.203.583)"
        subject: "Rijksmuseum"
        object: "KvK 41.203.583"
      note: |
        Links to formal registration identifiers (KvK, ISIL, etc.).
        The identifier should include the scheme and value.        
    
    - name: "has_legal_name"
      subject_type: "ORG.HER"
      predicate: "rov:legalName"
      object_type: "DEN"
      ontology: "rov"
      example:
        text: "Stichting Het Rijksmuseum (legally registered name)"
        subject: "Rijksmuseum"
        object: "Stichting Het Rijksmuseum"
      note: |
        rov:legalName is the registered legal name, which may differ
        from trading names or common usage names (skos:prefLabel).        
    
    - name: "has_org_status"
      subject_type: "ORG.HER"
      predicate: "rov:orgStatus"
      object_type: "skos:Concept"
      ontology: "rov"
      example:
        text: "The museum is currently active"
        subject: "The museum"
        object: "active"
      note: "Organization status: active, dissolved, merged, etc."
    
    # -------------------------------------------------------------------------
    # FOAF Relationship Patterns
    # -------------------------------------------------------------------------
    
    - name: "has_social_account"
      subject_type: "ORG.HER"
      predicate: "foaf:account"
      object_type: "TXT.SOC"
      ontology: "foaf"
      example:
        text: "Follow @rijksmuseum on Twitter"
        subject: "Rijksmuseum"
        object: "@rijksmuseum"
      note: |
        foaf:account links to foaf:OnlineAccount. More structured than
        schema:sameAs, allowing account name and service to be specified.        
    
    - name: "has_homepage"
      subject_type: "ORG.HER"
      predicate: "foaf:homepage"
      object_type: "TXT.IWB"
      ontology: "foaf"
      example:
        subject: "Rijksmuseum"
        object: "https://www.rijksmuseum.nl"
      note: "Alternative to schema:url. foaf:homepage is specifically the main website."

# =============================================================================
# SECTION 6: PiCO INTEGRATION PATTERNS
# =============================================================================
# Person in Context Ontology patterns for person observations.

pico_integration:

  description: |
    PiCO (Person in Context Ontology) models textual observations of persons
    as distinct from reconstructed person entities. This enables:
    - Tracking provenance of person mentions
    - Handling name variations across sources
    - Linking observations to formal person records    
  
  observation_pattern:
    description: "Every person mention creates a PersonObservation"
    class: "picom:PersonObservation"
    properties:
      - property: "picom:hasObservedName"
        description: "The name string as it appears in text"
        range: "pnv:PersonName"
      - property: "picom:isObservationOf"
        description: "Links to reconstructed Person entity"
        range: "crm:E21_Person"
      - property: "prov:hadPrimarySource"
        description: "The source document/webpage"
        range: "prov:Entity"
      - property: "picom:observedAt"
        description: "When the observation was made"
        range: "xsd:dateTime"
  
  pnv_name_structure:
    description: "Person Name Vocabulary for structured names"
    class: "pnv:PersonName"
    components:
      - property: "pnv:givenName"
        examples:
          - "Rembrandt"
          - "Maria"
          - "Jan"
      - property: "pnv:patronym"
        examples:
          - "Harmenszoon"
          - "Janszoon"
      - property: "pnv:surnamePrefix"
        examples:
          - "van"
          - "de"
          - "van den"
          - "op de"
      - property: "pnv:baseSurname"
        examples:
          - "Rijn"
          - "Berg"
          - "Velde"
      - property: "pnv:honorificPrefix"
        examples:
          - "Dr."
          - "Prof."
          - "Sir"
          - "Queen"
      - property: "pnv:honorificSuffix"
        examples:
          - "PhD"
          - "Jr."
          - "III"
  
  example:
    text: "Dr. Maria van den Berg, Director"
    observation:
      type: "picom:PersonObservation"
      hasObservedName:
        type: "pnv:PersonName"
        honorificPrefix: "Dr."
        givenName: "Maria"
        surnamePrefix: "van den"
        baseSurname: "Berg"
      hasRole: "Director"
      hadPrimarySource: "https://example.org/source-page"

# =============================================================================
# SECTION 7: NIF/NERD/OA INTEGRATION PATTERNS
# =============================================================================
# NLP Interchange Format (NIF), Named Entity Recognition and Disambiguation
# (NERD), and W3C Web Annotation (OA) patterns for cross-tool interoperability.

nif_nerd_integration:

  description: |
    This section defines how GLAM-NER annotations integrate with:
    - NIF 2.0: Standard format for NLP tool interchange (string/offset addressing)
    - NERD: Cross-system entity type mappings (10 core classes)
    - W3C OA: Web Annotation Data Model for annotation provenance
    - itsrdf: ITS 2.0 entity linking predicates
    
    These standards enable GLAM-NER annotations to be consumed by external
    NLP pipelines, linked data systems, and annotation aggregators.    

  nif_core_patterns:
    description: |
      NIF (NLP Interchange Format) provides URI-based addressing for text
      spans. Every annotation creates a nif:Phrase linked to its context.      
    
    context_class: "nif:Context"
    context_note: |
      nif:Context represents the full text document. All annotations reference
      this context via nif:referenceContext.      
    
    phrase_class: "nif:Phrase"
    phrase_note: |
      nif:Phrase represents extracted entity mentions. Each GLAM-NER entity
      becomes a nif:Phrase with offset-based URI addressing.      
    
    uri_schemes:
      offset_based:
        pattern: "{source_url}#offset_{begin}_{end}"
        example: "https://example.org/page#offset_42_58"
        note: "Default scheme. Begin/end are character offsets (0-based)."
      
      rfc5147:
        pattern: "{source_url}#char={begin},{end}"
        example: "https://example.org/page#char=42,58"
        note: "RFC 5147 fragment identifiers for text/plain."
      
      context_hash:
        pattern: "{source_url}#hash_{context_length}_{hash}_{begin}_{end}"
        note: "Hash-based URIs for content-addressing. More stable across edits."
    
    core_properties:
      - property: "nif:beginIndex"
        range: "xsd:nonNegativeInteger"
        description: "Character offset where entity span begins (0-based)"
      
      - property: "nif:endIndex"
        range: "xsd:nonNegativeInteger"
        description: "Character offset where entity span ends (exclusive)"
      
      - property: "nif:anchorOf"
        range: "xsd:string"
        description: "The exact text string of the entity mention"
      
      - property: "nif:referenceContext"
        range: "nif:Context"
        description: "Link to the document context containing this phrase"
      
      - property: "nif:sourceUrl"
        range: "xsd:anyURI"
        description: "Original URL of the source document"

  nerd_class_mappings:
    description: |
      NERD (Named Entity Recognition and Disambiguation) defines 10 core
      entity classes that map across multiple NER systems (DBpedia Spotlight,
      AlchemyAPI, OpenCalais, Zemanta, etc.). GLAM-NER types map to NERD for
      cross-system interoperability.      
    
    core_classes:
      - nerd_class: "nerd:Thing"
        description: "Generic entity (base class)"
        glam_ner_types: ["THING"]
      
      - nerd_class: "nerd:Person"
        description: "Human beings"
        glam_ner_types: ["BEING.PER", "BEING.STF"]
        subclasses:
          - "nerd:Astronaut"
          - "nerd:Politician"
          - "nerd:Artist"
      
      - nerd_class: "nerd:Organization"
        description: "Organizations, companies, institutions"
        glam_ner_types: ["ORG", "ORG.HER", "ORG.COM", "ORG.GOV", "ORG.EDU", "ORG.REL", "ORG.UNT"]
        subclasses:
          - "nerd:Company"
          - "nerd:SportsTeam"
          - "nerd:Band"
          - "nerd:University"
          - "nerd:Museum"
      
      - nerd_class: "nerd:Location"
        description: "Geographic places and features"
        glam_ner_types: ["PLACE", "PLACE.GEO", "PLACE.BLD", "PLACE.ADR"]
        subclasses:
          - "nerd:City"
          - "nerd:Country"
          - "nerd:Continent"
          - "nerd:Region"
          - "nerd:Facility"
      
      - nerd_class: "nerd:Event"
        description: "Named events"
        glam_ner_types: ["THING.EVT"]
        subclasses:
          - "nerd:SportEvent"
          - "nerd:MusicFestival"
      
      - nerd_class: "nerd:Time"
        description: "Temporal expressions"
        glam_ner_types: ["TEMPORAL", "TEMPORAL.DAT", "TEMPORAL.PER", "TEMPORAL.DUR"]
      
      - nerd_class: "nerd:Amount"
        description: "Quantities and measurements"
        glam_ner_types: ["QUANTITY", "QUANTITY.MON", "QUANTITY.CNT", "QUANTITY.DIM"]
      
      - nerd_class: "nerd:Product"
        description: "Products and creative works"
        glam_ner_types: ["THING.OBJ", "THING.DOC", "THING.COL"]
        subclasses:
          - "nerd:Album"
          - "nerd:Book"
          - "nerd:Movie"
          - "nerd:Software"
      
      - nerd_class: "nerd:Animal"
        description: "Animals"
        glam_ner_types: ["BEING.ANI"]
      
      - nerd_class: "nerd:Function"
        description: "Roles, titles, occupations"
        glam_ner_types: []
        note: "GLAM-NER treats roles as attributes of BEING.STF, not separate entities"

  web_annotation_patterns:
    description: |
      W3C Web Annotation Data Model provides standard annotation structure
      with target selectors for precise text span identification.      
    
    annotation_class: "oa:Annotation"
    
    structure:
      - property: "oa:hasBody"
        description: "The annotation content (entity type, confidence)"
        example: "The GLAM-NER entity classification"
      
      - property: "oa:hasTarget"
        description: "What is being annotated (text span)"
        example: "TextPositionSelector pointing to entity mention"
      
      - property: "oa:motivatedBy"
        description: "Why the annotation was created"
        value: "oa:classifying"
        note: "NER annotations are classification activities"
    
    target_selectors:
      text_position:
        class: "oa:TextPositionSelector"
        properties:
          - "oa:start"  # Start offset (0-based)
          - "oa:end"    # End offset (exclusive)
        note: "Equivalent to NIF offset-based addressing"
      
      text_quote:
        class: "oa:TextQuoteSelector"
        properties:
          - "oa:exact"   # The exact matched text
          - "oa:prefix"  # Context before (for disambiguation)
          - "oa:suffix"  # Context after (for disambiguation)
        note: "Provides context for robust text matching"
      
      xpath:
        class: "oa:XPathSelector"
        properties:
          - "rdf:value"  # XPath expression to DOM node
        note: "For HTML/XML sources with DOM structure"

  itsrdf_entity_linking:
    description: |
      ITS 2.0 (Internationalization Tag Set) provides entity linking predicates
      for connecting mentions to knowledge bases.      
    
    properties:
      - property: "itsrdf:taIdentRef"
        description: "URI reference to entity in knowledge base"
        example: "http://dbpedia.org/resource/Rijksmuseum"
        note: "Primary entity linking predicate"
      
      - property: "itsrdf:taSource"
        description: "Knowledge base source"
        examples:
          - "DBpedia"
          - "Wikidata"
          - "GeoNames"
      
      - property: "itsrdf:taConfidence"
        description: "Linking confidence score (0.0-1.0)"
        range: "xsd:double"
      
      - property: "itsrdf:taClassRef"
        description: "URI of entity type in target ontology"
        example: "http://dbpedia.org/ontology/Museum"

  glam_ner_to_nerd_mapping:
    description: "Complete mapping table from GLAM-NER types to NERD classes"
    mappings:
      - glam_type: "BEING"
        glam_code: "BEI"
        nerd_class: "nerd:Person"
        nerd_subclasses: []
      
      - glam_type: "BEING.PER"
        glam_code: "BEI.PER"
        nerd_class: "nerd:Person"
        nerd_subclasses: ["nerd:Artist", "nerd:Politician", "nerd:Astronaut"]
      
      - glam_type: "BEING.STF"
        glam_code: "BEI.STF"
        nerd_class: "nerd:Person"
        note: "Staff roles map to nerd:Function as secondary annotation"
      
      - glam_type: "BEING.ANI"
        glam_code: "BEI.ANI"
        nerd_class: "nerd:Animal"
      
      - glam_type: "BEING.MYT"
        glam_code: "BEI.MYT"
        nerd_class: "nerd:Person"
        note: "Mythological figures treated as Person in NERD"
      
      - glam_type: "BEING.GRP"
        glam_code: "BEI.GRP"
        nerd_class: "nerd:Organization"
        note: "Named groups map to Organization"
      
      - glam_type: "ORG"
        glam_code: "ORG"
        nerd_class: "nerd:Organization"
      
      - glam_type: "ORG.HER"
        glam_code: "ORG.HER"
        nerd_class: "nerd:Organization"
        nerd_subclasses: ["nerd:Museum"]
      
      - glam_type: "ORG.COM"
        glam_code: "ORG.COM"
        nerd_class: "nerd:Organization"
        nerd_subclasses: ["nerd:Company"]
      
      - glam_type: "ORG.GOV"
        glam_code: "ORG.GOV"
        nerd_class: "nerd:Organization"
      
      - glam_type: "ORG.EDU"
        glam_code: "ORG.EDU"
        nerd_class: "nerd:Organization"
        nerd_subclasses: ["nerd:University"]
      
      - glam_type: "ORG.REL"
        glam_code: "ORG.REL"
        nerd_class: "nerd:Organization"
      
      - glam_type: "ORG.UNT"
        glam_code: "ORG.UNT"
        nerd_class: "nerd:Organization"
        note: "Organizational units are Organizations in NERD"
      
      - glam_type: "PLACE"
        glam_code: "PLA"
        nerd_class: "nerd:Location"
      
      - glam_type: "PLACE.GEO"
        glam_code: "PLA.GEO"
        nerd_class: "nerd:Location"
        nerd_subclasses: ["nerd:City", "nerd:Country", "nerd:Region", "nerd:Continent"]
      
      - glam_type: "PLACE.BLD"
        glam_code: "PLA.BLD"
        nerd_class: "nerd:Location"
        nerd_subclasses: ["nerd:Facility"]
      
      - glam_type: "PLACE.ADR"
        glam_code: "PLA.ADR"
        nerd_class: "nerd:Location"
      
      - glam_type: "TEMPORAL"
        glam_code: "TMP"
        nerd_class: "nerd:Time"
      
      - glam_type: "TEMPORAL.DAT"
        glam_code: "TMP.DAT"
        nerd_class: "nerd:Time"
      
      - glam_type: "TEMPORAL.PER"
        glam_code: "TMP.PER"
        nerd_class: "nerd:Time"
      
      - glam_type: "TEMPORAL.DUR"
        glam_code: "TMP.DUR"
        nerd_class: "nerd:Time"
      
      - glam_type: "QUANTITY"
        glam_code: "QTY"
        nerd_class: "nerd:Amount"
      
      - glam_type: "QUANTITY.MON"
        glam_code: "QTY.MON"
        nerd_class: "nerd:Amount"
      
      - glam_type: "QUANTITY.CNT"
        glam_code: "QTY.CNT"
        nerd_class: "nerd:Amount"
      
      - glam_type: "QUANTITY.DIM"
        glam_code: "QTY.DIM"
        nerd_class: "nerd:Amount"
      
      - glam_type: "THING"
        glam_code: "THI"
        nerd_class: "nerd:Thing"
      
      - glam_type: "THING.OBJ"
        glam_code: "THI.OBJ"
        nerd_class: "nerd:Product"
      
      - glam_type: "THING.DOC"
        glam_code: "THI.DOC"
        nerd_class: "nerd:Product"
        nerd_subclasses: ["nerd:Book"]
      
      - glam_type: "THING.COL"
        glam_code: "THI.COL"
        nerd_class: "nerd:Thing"
        note: "Collections map to generic Thing (no NERD equivalent)"
      
      - glam_type: "THING.EVT"
        glam_code: "THI.EVT"
        nerd_class: "nerd:Event"
        nerd_subclasses: ["nerd:SportEvent", "nerd:MusicFestival"]
      
      - glam_type: "TXT"
        glam_code: "TXT"
        nerd_class: "nerd:Thing"
        note: "Text fragments have no direct NERD mapping"

  example_nif_annotation:
    description: "Complete example of GLAM-NER annotation in NIF/OA format"
    source_text: "The Rijksmuseum in Amsterdam holds over one million objects."
    
    turtle_example: |
      @prefix nif: <http://persistence.uni-leipzig.org/nlp2rdf/ontologies/nif-core#> .
      @prefix nerd: <http://nerd.eurecom.fr/ontology#> .
      @prefix oa: <http://www.w3.org/ns/oa#> .
      @prefix itsrdf: <http://www.w3.org/2005/11/its/rdf#> .
      @prefix glam: <https://w3id.org/glam/> .
      @prefix xsd: <http://www.w3.org/2001/XMLSchema#> .
      
      # Document context
      <https://example.org/doc1#offset_0_60> a nif:Context, nif:OffsetBasedString ;
          nif:isString "The Rijksmuseum in Amsterdam holds over one million objects." ;
          nif:sourceUrl <https://example.org/doc1> ;
          nif:beginIndex "0"^^xsd:nonNegativeInteger ;
          nif:endIndex "60"^^xsd:nonNegativeInteger .
      
      # Entity 1: Rijksmuseum (ORG.HER)
      <https://example.org/doc1#offset_4_15> a nif:Phrase, nif:OffsetBasedString ;
          nif:anchorOf "Rijksmuseum" ;
          nif:beginIndex "4"^^xsd:nonNegativeInteger ;
          nif:endIndex "15"^^xsd:nonNegativeInteger ;
          nif:referenceContext <https://example.org/doc1#offset_0_60> ;
          a nerd:Organization, nerd:Museum ;
          glam:entityType "ORG.HER" ;
          glam:confidence "0.95"^^xsd:double ;
          itsrdf:taIdentRef <http://www.wikidata.org/entity/Q190804> ;
          itsrdf:taSource "Wikidata" ;
          itsrdf:taConfidence "0.92"^^xsd:double .
      
      # Entity 2: Amsterdam (PLACE.GEO)
      <https://example.org/doc1#offset_19_28> a nif:Phrase, nif:OffsetBasedString ;
          nif:anchorOf "Amsterdam" ;
          nif:beginIndex "19"^^xsd:nonNegativeInteger ;
          nif:endIndex "28"^^xsd:nonNegativeInteger ;
          nif:referenceContext <https://example.org/doc1#offset_0_60> ;
          a nerd:Location, nerd:City ;
          glam:entityType "PLACE.GEO" ;
          glam:confidence "0.98"^^xsd:double ;
          itsrdf:taIdentRef <http://www.wikidata.org/entity/Q727> ;
          itsrdf:taSource "Wikidata" .
      
      # Entity 3: one million (QUANTITY.CNT)
      <https://example.org/doc1#offset_41_52> a nif:Phrase, nif:OffsetBasedString ;
          nif:anchorOf "one million" ;
          nif:beginIndex "41"^^xsd:nonNegativeInteger ;
          nif:endIndex "52"^^xsd:nonNegativeInteger ;
          nif:referenceContext <https://example.org/doc1#offset_0_60> ;
          a nerd:Amount ;
          glam:entityType "QUANTITY.CNT" ;
          glam:normalizedValue "1000000"^^xsd:integer ;
          glam:confidence "0.90"^^xsd:double .      

# =============================================================================
# SECTION 8: PROPERTY EXTRACTION RULES
# =============================================================================
# Rules for extracting relationships and attributes between entities.
# Properties connect entities to each other and to literal values.

property_extraction:

  description: |
    Property extraction identifies relationships between detected entities
    and extracts attributes with their values. Every property claim MUST
    include verifiable provenance.
    
    Properties are categorized by:
    - Subject entity type (e.g., ORG.HER, BEING.PER)
    - Property semantics (temporal, spatial, organizational, etc.)
    - Value type (entity reference, literal, temporal range)
    
    Each extracted property becomes a claim that requires:
    1. Source path (XPath, JSONPath, or text offset)
    2. Extraction timestamp
    3. Agent identifier
    4. Convention version reference    

  property_categories:
    
    temporal_properties:
      description: "Properties with date/time values or temporal ranges"
      
      founding_date:
        property_uri: "schema:foundingDate"
        alternative_uris:
          - "crm:P98i_was_born"
          - "rico:beginningDate"
        subject_types: ["ORG", "ORG.HER", "ORG.COM", "ORG.GOV"]
        value_type: "temporal_expression"
        patterns:
          - pattern: "founded in {YEAR}"
            example: "The museum was founded in 1885"
          - pattern: "established {YEAR}"
            example: "Established 1952"
          - pattern: "since {YEAR}"
            example: "Serving the community since 1923"
          - pattern: "opgericht in {YEAR}"
            example: "Opgericht in 1901"
          - pattern: "{YEAR} gesticht"
            example: "In 1850 gesticht"
        extraction_notes: |
          Temporal expressions may be:
          - Precise dates: "15 July 1885"
          - Year only: "1885"
          - Decade: "1880s"
          - Century: "19th century"
          - Relative: "over 100 years ago"
          
          Normalize to ISO 8601 where possible, preserve original text.          
      
      dissolution_date:
        property_uri: "schema:dissolutionDate"
        alternative_uris:
          - "crm:P100i_died_in"
          - "rico:endDate"
        subject_types: ["ORG", "ORG.HER"]
        value_type: "temporal_expression"
        patterns:
          - pattern: "closed in {YEAR}"
          - pattern: "dissolved {YEAR}"
          - pattern: "ceased operations {YEAR}"
      
      opening_hours:
        property_uri: "schema:openingHours"
        subject_types: ["ORG.HER", "PLC.BLD"]
        value_type: "temporal_specification"
        patterns:
          - pattern: "open {DAYS} {TIME_RANGE}"
            example: "Open Tuesday-Sunday 10:00-17:00"
          - pattern: "geopend {DAYS} {TIME_RANGE}"
            example: "Geopend dinsdag t/m zondag van 10:00 tot 17:00"
      
      event_date:
        property_uri: "schema:startDate"
        alternative_uris:
          - "crm:P4_has_time-span"
        subject_types: ["THG.EVT", "DEN.EXH"]
        value_type: "temporal_expression"
    
    organizational_properties:
      description: "Properties describing organizational relationships"
      
      parent_organization:
        property_uri: "org:subOrganizationOf"
        alternative_uris:
          - "schema:parentOrganization"
          - "rico:isOrWasSubordinateTo"
        subject_types: ["ORG", "ORG.HER", "ORG.UNT"]
        value_type: "entity_reference"
        target_types: ["ORG", "ORG.PAR", "ORG.GOV"]
        patterns:
          - pattern: "part of {ORG}"
            example: "The archive is part of the National Library"
          - pattern: "department of {ORG}"
          - pattern: "onderdeel van {ORG}"
          - pattern: "valt onder {ORG}"
      
      has_unit:
        property_uri: "org:hasUnit"
        alternative_uris:
          - "schema:department"
        subject_types: ["ORG", "ORG.HER"]
        value_type: "entity_reference"
        target_types: ["ORG.UNT"]
        patterns:
          - pattern: "{ORG} has a {UNIT}"
          - pattern: "includes the {UNIT}"
      
      member_of:
        property_uri: "org:memberOf"
        alternative_uris:
          - "schema:memberOf"
        subject_types: ["ORG.HER", "BEING.STF"]
        value_type: "entity_reference"
        target_types: ["ORG", "ORG.HER"]
      
      has_member:
        property_uri: "org:hasMember"
        subject_types: ["ORG", "ORG.HER"]
        value_type: "entity_reference"
        target_types: ["BEING", "BEING.STF", "ORG.HER"]
    
    contact_properties:
      description: "Properties for contact information"
      
      email:
        property_uri: "schema:email"
        alternative_uris:
          - "vcard:hasEmail"
        subject_types: ["ORG.HER", "BEING.STF"]
        value_type: "email_address"
        patterns:
          - pattern: "{EMAIL_REGEX}"
            regex: "[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\\.[a-zA-Z]{2,}"
      
      telephone:
        property_uri: "schema:telephone"
        alternative_uris:
          - "vcard:hasTelephone"
        subject_types: ["ORG.HER"]
        value_type: "phone_number"
        patterns:
          - pattern: "+{COUNTRY_CODE} {NUMBER}"
          - pattern: "0{AREA_CODE}-{NUMBER}"
          - pattern: "({AREA_CODE}) {NUMBER}"
      
      address:
        property_uri: "schema:address"
        alternative_uris:
          - "vcard:hasAddress"
          - "org:hasSite"
        subject_types: ["ORG.HER"]
        value_type: "entity_reference"
        target_types: ["PLC.ADR", "PLC.IAD"]
      
      location:
        property_uri: "schema:location"
        alternative_uris:
          - "crm:P55_has_current_location"
          - "rico:hasOrHadLocation"
        subject_types: ["ORG.HER", "THG.EVT"]
        value_type: "entity_reference"
        target_types: ["PLC", "PLC.GEO", "PLC.BLD"]
    
    identifier_properties:
      description: "Properties for formal identifiers"
      
      isil_code:
        property_uri: "glam:isilCode"
        alternative_uris:
          - "schema:identifier"
        subject_types: ["ORG.HER"]
        value_type: "identifier"
        pattern: "[A-Z]{2}-[A-Za-z0-9]+"
        example: "NL-AsdRM"
      
      wikidata_id:
        property_uri: "schema:sameAs"
        subject_types: ["ORG.HER", "BEING", "PLC"]
        value_type: "uri"
        pattern: "Q[0-9]+"
        uri_template: "http://www.wikidata.org/entity/{ID}"
      
      kvk_number:
        property_uri: "rov:registration"
        subject_types: ["ORG.HER", "ORG.COM"]
        value_type: "identifier"
        pattern: "[0-9]{8}"
        jurisdiction: "NL"
    
    collection_properties:
      description: "Properties for collection descriptions"
      
      collection_size:
        property_uri: "schema:size"
        alternative_uris:
          - "rico:recordSetExtent"
        subject_types: ["ORG.HER", "DEN.COL"]
        value_type: "quantity"
        patterns:
          - pattern: "{NUMBER} {UNIT}"
            examples:
              - "over 1 million objects"
              - "8,000 artworks"
              - "15 kilometers of archives"
      
      subject_area:
        property_uri: "schema:about"
        alternative_uris:
          - "dct:subject"
        subject_types: ["ORG.HER", "DEN.COL"]
        value_type: "concept"
      
      temporal_coverage:
        property_uri: "dct:temporal"
        subject_types: ["DEN.COL", "ORG.HER"]
        value_type: "temporal_range"
        patterns:
          - pattern: "from {START} to {END}"
          - pattern: "{CENTURY} century"
          - pattern: "{ERA} period"

  extraction_pipeline:
    description: "Steps for property extraction"
    
    steps:
      - step: 1
        name: "Entity Detection"
        description: "First identify all entities in the text"
        output: "List of entity mentions with spans and types"
      
      - step: 2
        name: "Relation Pattern Matching"
        description: "Apply relation patterns to detect property claims"
        input: "Entity pairs and intervening text"
        output: "Candidate property claims"
      
      - step: 3
        name: "Property Classification"
        description: "Classify property type using ontology mappings"
        input: "Candidate claims"
        output: "Typed property claims with URIs"
      
      - step: 4
        name: "Value Normalization"
        description: "Normalize values (dates, quantities, identifiers)"
        input: "Raw extracted values"
        output: "Normalized values with original text preserved"
      
      - step: 5
        name: "Provenance Attachment"
        description: "Add required provenance metadata to each claim"
        input: "Property claims"
        output: "Claims with full provenance (see claim_validation section)"

# =============================================================================
# SECTION 9: CLAIM VALIDATION SCHEMA
# =============================================================================
# LinkML-based schema for claim validation and provenance tracking.
# Every extracted claim (entity or property) MUST conform to this schema.

claim_validation:

  description: |
    All claims extracted by this convention MUST include complete provenance
    metadata. This enables verification, reproducibility, and trust assessment.
    
    A "claim" is any assertion extracted from a source:
    - Entity existence ("Rijksmuseum" is an organization)
    - Property value ("Rijksmuseum" was founded in 1885)
    - Relationship ("Dr. Dibbits" is director of "Rijksmuseum")
    
    Claims without verifiable provenance are considered FABRICATED and
    MUST be rejected or flagged for manual verification.    

  linkml_schema_reference:
    schema_file: "schemas/20251121/linkml/modules/classes/WebClaim.yaml"
    namespace: "https://nde.nl/ontology/hc/"
    description: |
      The WebClaim class defines the formal structure for web-extracted claims.
      All claim instances MUST validate against this LinkML schema.      

  required_provenance_components:
    description: |
      Every claim MUST include these 5 provenance components.
      Missing any component renders the claim unverifiable.      
    
    components:
      - id: 1
        name: "namespace"
        description: "Ontology namespace for the claim predicate"
        examples:
          - "schema:" 
          - "rico:"
          - "org:"
          - "crm:"
        linkml_slot: "claim_type"
        required: true
      
      - id: 2
        name: "path"
        description: "Location path to the claim source in the document"
        path_types:
          xpath:
            description: "XPath expression for HTML/XML documents"
            pattern: "^/.*"
            example: "/html[1]/body[1]/div[6]/div[1]/h1[1]"
          jsonpath:
            description: "JSONPath expression for JSON documents"
            pattern: "^\\$.*"
            example: "$.organization.foundingDate"
          text_offset:
            description: "Character offset range for plain text"
            pattern: "offset_{begin}_{end}"
            example: "offset_42_58"
          css_selector:
            description: "CSS selector for HTML elements"
            example: "footer > div.contact > a.email"
        linkml_slot: "xpath"
        required: true
      
      - id: 3
        name: "timestamp"
        description: "When the source was retrieved/observed"
        format: "ISO 8601 with timezone"
        example: "2025-12-02T14:30:00Z"
        linkml_slot: "retrieved_on"
        required: true
      
      - id: 4
        name: "agent"
        description: "The extraction agent (model, script, human)"
        agent_types:
          llm_model:
            examples:
              - "glm4.6"
              - "gpt-4-turbo"
              - "claude-3.5-sonnet"
            format: "{model_family}-{version}"
          script:
            examples:
              - "add_xpath_provenance.py"
              - "extract_entities.py"
            format: "{script_name}"
          human:
            examples:
              - "curator:john.doe@museum.nl"
            format: "human:{identifier}"
        linkml_slot: "extraction_method"
        required: true
      
      - id: 5
        name: "context_convention"
        description: "Version of this convention used for extraction"
        format: "entity_annotation_rules_v{VERSION}"
        current_value: "entity_annotation_rules_v1.6.3"
        linkml_slot: "extraction_context"
        required: true

  claim_structure:
    description: "Complete claim structure conforming to LinkML WebClaim schema"
    
    claim_fields:
      # Identity
      claim_id:
        type: "uriorcurie"
        description: "Unique identifier for this claim"
        pattern: "urn:glam:claim:{uuid}"
        required: true
      
      # Claim content
      claim_type:
        type: "ClaimTypeEnum"
        description: "Type of claim (from CanonicalClaimTypes enum)"
        required: true
      
      claim_value:
        type: "string"
        description: "The extracted value"
        required: true
      
      normalized_value:
        type: "varies"
        description: "Normalized/parsed value (dates, quantities)"
        required: false
      
      # Subject and object entities (for relationship claims)
      subject_entity_id:
        type: "uriorcurie"
        description: "Entity this claim is about"
        required: true
      
      object_entity_id:
        type: "uriorcurie"
        description: "Target entity (for relationship claims)"
        required: false
      
      # Provenance (all 5 components)
      source_url:
        type: "uri"
        description: "URL of source document"
        required: true
      
      xpath:
        type: "string"
        description: "Path to claim in source document"
        required: true
      
      html_file:
        type: "string"
        description: "Path to archived source file"
        required: true
      
      retrieved_on:
        type: "datetime"
        description: "When source was retrieved"
        required: true
      
      extraction_agent:
        type: "string"
        description: "Agent that extracted the claim"
        required: true
      
      extraction_convention:
        type: "string"
        description: "Convention version used"
        default: "entity_annotation_rules_v1.6.3"
        required: true
      
      # Verification
      xpath_match_score:
        type: "float"
        range: "0.0-1.0"
        description: "Match quality between claim and source text"
        required: true
      
      xpath_matched_text:
        type: "string"
        description: "Actual text found at xpath location"
        required: false
      
      # Confidence (computed, not estimated)
      confidence:
        type: "float"
        range: "0.0-1.0"
        description: "Computed confidence based on xpath_match_score and extraction_method"
        computation: |
          confidence = xpath_match_score * method_weight
          where method_weight:
            - xpath_exact_match: 1.0
            - pattern_match: 0.9
            - nlp_extraction: 0.7
            - llm_extraction: 0.6          
        required: true

  claim_tiers:
    description: "Claims are categorized by reliability tier"
    
    tiers:
      - tier: 1
        name: "STRUCTURAL"
        description: "Extracted from HTML structure with exact XPath"
        confidence_range: "0.95-1.0"
        examples:
          - "page_title from <title> tag"
          - "email from mailto: link href"
          - "isil_code from structured data"
        verification: "Automated - xpath resolves to exact text"
      
      - tier: 2
        name: "PATTERN"
        description: "Extracted via pattern matching with XPath"
        confidence_range: "0.80-0.95"
        examples:
          - "phone number matching +31 pattern"
          - "founding year from 'founded in YYYY' pattern"
        verification: "Automated with pattern validation"
      
      - tier: 3
        name: "NLP"
        description: "Extracted via NLP/LLM with XPath verification"
        confidence_range: "0.50-0.80"
        examples:
          - "organization description from about page"
          - "staff member name from team page"
        verification: "Requires XPath verification, may need human review"
      
      - tier: 4
        name: "INFERRED"
        description: "Inferred from multiple claims, not directly extracted"
        confidence_range: "0.30-0.50"
        examples:
          - "organization type inferred from name patterns"
          - "relationship inferred from context"
        verification: "Requires human review"

  validation_rules:
    description: "Rules for validating claim instances"
    
    rules:
      - id: "CV001"
        rule: "Every claim MUST have all 5 provenance components"
        severity: "ERROR"
        check: "xpath AND retrieved_on AND extraction_agent AND extraction_convention AND source_url"
      
      - id: "CV002"
        rule: "xpath_match_score MUST be computed, not estimated"
        severity: "ERROR"
        check: "xpath_match_score derived from text comparison"
      
      - id: "CV003"
        rule: "Claims with xpath_match_score < 0.3 MUST be flagged"
        severity: "WARNING"
        action: "Add to manual_review queue"
      
      - id: "CV004"
        rule: "Temporal claims MUST include normalized ISO 8601 value"
        severity: "WARNING"
        applies_to: ["founding_date", "dissolution_date", "event_date"]
      
      - id: "CV005"
        rule: "Entity reference claims MUST include target entity_id"
        severity: "ERROR"
        applies_to: ["parent_organization", "location", "member_of"]
      
      - id: "CV006"
        rule: "Extraction convention version MUST match current production version"
        severity: "WARNING"
        current_version: "entity_annotation_rules_v1.6.3"

  example_validated_claim:
    description: "Complete example of a validated property claim"
    
    claim:
      claim_id: "urn:glam:claim:550e8400-e29b-41d4-a716-446655440000"
      claim_type: "founding_date"
      claim_value: "1885"
      normalized_value: "1885-01-01"
      
      subject_entity_id: "urn:glam:entity:rijksmuseum"
      
      # Provenance (5 components)
      source_url: "https://www.rijksmuseum.nl/nl/over-ons/wat-we-doen/geschiedenis"
      xpath: "/html[1]/body[1]/main[1]/article[1]/div[2]/p[3]"
      html_file: "web/0001/rijksmuseum.nl/over-ons/geschiedenis.html"
      retrieved_on: "2025-12-02T10:30:00Z"
      extraction_agent: "glm4.6"
      extraction_convention: "entity_annotation_rules_v1.6.3"
      
      # Verification
      xpath_match_score: 0.92
      xpath_matched_text: "Het Rijksmuseum is in 1885 opgericht als 's Rijks Museum"
      
      # Computed confidence
      confidence: 0.644  # 0.92 * 0.7 (NLP extraction weight)
      tier: 3

# =============================================================================
# SECTION 10: ENTITY RESOLUTION AND LINKING PIPELINE
# =============================================================================
# Complete pipeline for entity resolution (disambiguation) and entity linking
# (connecting to knowledge bases).

entity_resolution_linking:

  description: |
    After entity recognition and property extraction, entities must be:
    1. RESOLVED: Disambiguated and merged (same entity, different mentions)
    2. LINKED: Connected to authoritative knowledge bases (Wikidata, ISIL, etc.)
    
    Both processes generate claims that require provenance tracking.    

  entity_resolution:
    description: |
      Entity resolution determines when two entity mentions refer to the
      same real-world entity. This is critical for:
      - Merging data from multiple sources
      - Preventing duplicate records
      - Building knowledge graphs      
    
    resolution_strategies:
      
      identifier_matching:
        description: "Exact match on formal identifiers"
        priority: 1
        confidence: 1.0
        identifiers:
          - type: "ISIL"
            pattern: "[A-Z]{2}-[A-Za-z0-9]+"
            example: "Both have ISIL NL-AsdRM → same entity"
          - type: "Wikidata"
            pattern: "Q[0-9]+"
            example: "Both link to Q190804 → same entity"
          - type: "KvK"
            pattern: "[0-9]{8}"
            example: "Both have KvK 12345678 → same entity"
      
      name_normalization:
        description: "Match after normalizing names"
        priority: 2
        confidence: 0.85
        normalization_steps:
          - step: "lowercase"
          - step: "remove_articles"
            articles: ["de", "het", "the", "a", "an", "le", "la", "der", "die", "das"]
          - step: "remove_legal_forms"
            forms: ["stichting", "vereniging", "foundation", "inc", "ltd", "bv", "nv"]
          - step: "normalize_whitespace"
          - step: "remove_diacritics"
        example: |
          "Stichting Rijksmuseum" → "rijksmuseum"
          "The Rijksmuseum" → "rijksmuseum"
          → Same normalized form, likely same entity          
      
      location_name_combo:
        description: "Match on (location, normalized_name) pair"
        priority: 3
        confidence: 0.75
        example: |
          Entity A: city="Amsterdam", name="City Archive"
          Entity B: city="Amsterdam", name="Stadsarchief"
          → Same location + similar function, possible match (needs review)          
      
      fuzzy_matching:
        description: "Similarity-based matching for name variants"
        priority: 4
        confidence: 0.60
        algorithms:
          - name: "Levenshtein"
            threshold: 0.85
          - name: "Jaro-Winkler"
            threshold: 0.90
          - name: "Token Set Ratio"
            threshold: 0.85
        example: |
          "Rijksmuseum Amsterdam" vs "Amsterdam Rijksmuseum"
          Token Set Ratio: 1.0 → likely match          

    resolution_output:
      description: "Resolution creates a merge claim with provenance"
      claim_type: "entity_resolution"
      structure:
        resolution_id: "urn:glam:resolution:{uuid}"
        entity_a_id: "urn:glam:entity:{id_a}"
        entity_b_id: "urn:glam:entity:{id_b}"
        resolution_strategy: "identifier_matching"
        resolution_confidence: 1.0
        resolved_entity_id: "urn:glam:entity:{canonical_id}"
        provenance:
          resolved_on: "2025-12-02T10:30:00Z"
          resolution_agent: "entity_resolver.py"
          resolution_convention: "entity_annotation_rules_v1.6.3"
          matching_evidence:
            - field: "isil_code"
              value_a: "NL-AsdRM"
              value_b: "NL-AsdRM"
              match_type: "exact"

  entity_linking:
    description: |
      Entity linking connects extracted entities to authoritative
      knowledge bases. This enables:
      - Enrichment with additional data
      - Cross-dataset integration
      - Semantic interoperability      
    
    target_knowledge_bases:
      
      wikidata:
        priority: 1
        namespace: "http://www.wikidata.org/entity/"
        lookup_methods:
          - method: "SPARQL search"
            description: "Query Wikidata SPARQL endpoint by name"
          - method: "API search"
            description: "Use wbsearchentities API"
          - method: "exact_match"
            description: "Direct P-property lookup (e.g., P791 for ISIL)"
        properties_to_check:
          - "P791"  # ISIL code
          - "P214"  # VIAF ID
          - "P244"  # Library of Congress authority ID
          - "P227"  # GND ID
        link_claim:
          property_uri: "schema:sameAs"
          alternative_uri: "owl:sameAs"
      
      isil_registry:
        priority: 2
        namespace: "https://ld.zdb-services.de/resource/"
        lookup_methods:
          - method: "ISIL code lookup"
            description: "Direct lookup by ISIL code"
        link_claim:
          property_uri: "schema:identifier"
      
      geonames:
        priority: 3
        namespace: "http://sws.geonames.org/"
        applies_to: ["PLC", "PLC.GEO", "PLC.SET"]
        lookup_methods:
          - method: "Coordinate lookup"
            description: "Find nearest named place by lat/lon"
          - method: "Name search"
            description: "Search by place name + country"
        link_claim:
          property_uri: "schema:geo"
      
      viaf:
        priority: 4
        namespace: "http://viaf.org/viaf/"
        applies_to: ["ORG.HER", "BEING"]
        lookup_methods:
          - method: "Name search"
            description: "Search VIAF by organization/person name"
        link_claim:
          property_uri: "schema:sameAs"

    linking_output:
      description: "Linking creates an entity link claim with provenance"
      claim_type: "entity_link"
      linkml_properties:
        - "itsrdf:taIdentRef"
        - "itsrdf:taSource"
        - "itsrdf:taConfidence"
      structure:
        link_id: "urn:glam:link:{uuid}"
        source_entity_id: "urn:glam:entity:{id}"
        target_uri: "http://www.wikidata.org/entity/Q190804"
        target_source: "Wikidata"
        link_confidence: 0.95
        link_method: "ISIL exact match via P791"
        provenance:
          linked_on: "2025-12-02T10:35:00Z"
          linking_agent: "entity_linker.py"
          linking_convention: "entity_annotation_rules_v1.6.3"
          matching_evidence:
            - source_field: "isil_code"
              source_value: "NL-AsdRM"
              target_property: "P791"
              target_value: "NL-AsdRM"
              match_type: "exact"

  pipeline_stages:
    description: "Complete extraction-to-linking pipeline"
    
    stages:
      - stage: 1
        name: "Source Acquisition"
        input: "URL or document"
        output: "Archived source with timestamp"
        claims_generated: ["source_retrieval"]
      
      - stage: 2
        name: "Entity Recognition"
        input: "Archived source"
        output: "Entity mentions with spans and types"
        claims_generated: ["entity_mention"]
        convention_section: "SECTION 2: ENTITY TYPE DEFINITIONS"
      
      - stage: 3
        name: "Property Extraction"
        input: "Entity mentions + source text"
        output: "Property claims linking entities"
        claims_generated: ["property_claim"]
        convention_section: "SECTION 8: PROPERTY EXTRACTION RULES"
      
      - stage: 4
        name: "Claim Validation"
        input: "All claims"
        output: "Validated claims with provenance"
        claims_generated: ["validated_claim"]
        convention_section: "SECTION 9: CLAIM VALIDATION SCHEMA"
      
      - stage: 5
        name: "Entity Resolution"
        input: "Validated entity claims"
        output: "Resolved entities (merged duplicates)"
        claims_generated: ["resolution_claim"]
        convention_section: "SECTION 10: ENTITY RESOLUTION AND LINKING"
      
      - stage: 6
        name: "Entity Linking"
        input: "Resolved entities"
        output: "Linked entities with KB references"
        claims_generated: ["link_claim"]
        convention_section: "SECTION 10: ENTITY RESOLUTION AND LINKING"
      
      - stage: 7
        name: "Output Generation"
        input: "All validated claims"
        output: "LinkML instances, RDF, JSON-LD"
        convention_section: "SECTION 11: EXTRACTION OUTPUT FORMAT"

# =============================================================================
# SECTION 11: EXTRACTION OUTPUT FORMAT
# =============================================================================
# Standard output format for NER extraction results.

output_format:

  description: |
    Extracted entities should be output in JSON format with the following
    structure. Each entity includes its span, type, confidence, and any
    relationships to other entities.
    
    The schema supports both native JSON output and NIF/RDF serialization.
    NIF properties are included to enable direct conversion to Turtle/JSON-LD
    for NLP tool interchange.    

  schema:
    document_context:
      type: "object"
      description: "Document-level context (maps to nif:Context)"
      properties:
        source_url:
          type: "string"
          format: "uri"
          nif_property: "nif:sourceUrl"
        full_text:
          type: "string"
          nif_property: "nif:isString"
        text_length:
          type: "integer"
          description: "Length of full_text for nif:endIndex"
    
    entities:
      type: "array"
      items:
        entity_id:
          type: "string"
          description: "Unique identifier for this entity in the extraction"
        text:
          type: "string"
          description: "The exact text span as it appears in source"
          nif_property: "nif:anchorOf"
        entity_type:
          type: "string"
          description: "Entity type code (e.g., ORG.HER, BEING.PER)"
          glam_property: "glam:entityType"
        start_offset:
          type: "integer"
          description: "Character offset where entity begins (0-based)"
          nif_property: "nif:beginIndex"
        end_offset:
          type: "integer"
          description: "Character offset where entity ends (exclusive)"
          nif_property: "nif:endIndex"
        confidence:
          type: "number"
          description: "Confidence score 0.0-1.0"
          glam_property: "glam:confidence"
        ontology_class:
          type: "string"
          description: "Primary ontology class URI"
        nerd_class:
          type: "string"
          description: "NERD ontology class for NLP interchange"
          example: "nerd:Organization"
        entity_link:
          type: "object"
          description: "Entity linking to knowledge bases (maps to itsrdf properties)"
          properties:
            reference_uri:
              type: "string"
              format: "uri"
              itsrdf_property: "itsrdf:taIdentRef"
              example: "http://www.wikidata.org/entity/Q190804"
            source:
              type: "string"
              itsrdf_property: "itsrdf:taSource"
              example: "Wikidata"
            linking_confidence:
              type: "number"
              itsrdf_property: "itsrdf:taConfidence"
        normalized_value:
          type: "string"
          description: "Normalized/standardized value (for dates, quantities)"
          glam_property: "glam:normalizedValue"
        attributes:
          type: "object"
          description: "Type-specific attributes (e.g., PNV components for names)"
        relationships:
          type: "array"
          description: "Relationships to other entities in extraction"
          items:
            predicate:
              type: "string"
            target_entity_id:
              type: "string"

  nif_uri_generation:
    description: "How to generate NIF-compliant URIs for entities"
    pattern: "{source_url}#offset_{start_offset}_{end_offset}"
    example: "https://example.org/page#offset_4_15"
    
  nif_serialization_note: |
    To convert JSON output to NIF Turtle:
    1. Create nif:Context from document_context
    2. For each entity, create nif:Phrase with offset-based URI
    3. Add nif:referenceContext linking to Context
    4. Add nerd: class assertions based on nerd_class field
    5. Add itsrdf: properties for entity linking    

  example:
    # DOCUMENT-LEVEL EXTRACTION CONTEXT
    # All 5 provenance components at document level
    document_context:
      source_url: "https://www.rijksmuseum.nl/en/about-us"
      full_text: "The Rijksmuseum is located at Museumstraat 1, 1071 XX Amsterdam. Contact: info@rijksmuseum.nl. Director: Dr. Taco Dibbits."
      text_length: 124
      
    # EXTRACTION PROVENANCE (5 components - document level)
    extraction_provenance:
      retrieved_on: "2025-12-02T10:30:00Z"           # Component 3: timestamp
      extraction_agent: "glm4.6"                     # Component 4: agent
      extraction_convention: "entity_annotation_rules_v1.6.3"  # Component 5: context
      html_archive_path: "web/0001/rijksmuseum.nl/en/about-us.html"
    
    # EXTRACTED ENTITIES
    entities:
      - entity_id: "e1"
        text: "Rijksmuseum"
        entity_type: "ORG.HER"
        start_offset: 4
        end_offset: 15
        confidence: 0.98
        ontology_class: "glam:HeritageCustodian"
        nerd_class: "nerd:Museum"
        
        # ENTITY PROVENANCE (5 components)
        provenance:
          namespace: "schema:"                       # Component 1: namespace
          xpath: "/html[1]/body[1]/main[1]/h1[1]"   # Component 2: path
          xpath_match_score: 1.0
          xpath_matched_text: "Rijksmuseum"
        
        entity_link:
          reference_uri: "http://www.wikidata.org/entity/Q190804"
          source: "Wikidata"
          linking_confidence: 0.96
        
        # PROPERTY CLAIMS with full 5-component provenance
        property_claims:
          - claim_id: "urn:glam:claim:e1-founding"
            claim_type: "founding_date"              # Component 1: namespace (rico:)
            claim_value: "1885"
            normalized_value: "1885-01-01"
            provenance:
              namespace: "rico:"
              xpath: "/html[1]/body[1]/main[1]/article[1]/section[2]/p[1]"
              retrieved_on: "2025-12-02T10:30:00Z"
              extraction_agent: "glm4.6"
              extraction_convention: "entity_annotation_rules_v1.6.3"
              html_file: "web/0001/rijksmuseum.nl/en/about-us.html"
              xpath_match_score: 0.85
              xpath_matched_text: "The Rijksmuseum was founded in 1885"
            confidence: 0.595  # 0.85 * 0.7 (NLP weight)
            tier: 3
        
        relationships:
          - predicate: "schema:address"
            target_entity_id: "e2"
          - predicate: "schema:email"
            target_entity_id: "e3"
      
      - entity_id: "e2"
        text: "Museumstraat 1, 1071 XX Amsterdam"
        entity_type: "PLC.IAD"
        start_offset: 30
        end_offset: 63
        confidence: 0.95
        ontology_class: "schema:PostalAddress"
        nerd_class: "nerd:Location"
        
        # ENTITY PROVENANCE (5 components)
        provenance:
          namespace: "schema:"
          xpath: "/html[1]/body[1]/main[1]/section[1]/address[1]"
          xpath_match_score: 1.0
          xpath_matched_text: "Museumstraat 1, 1071 XX Amsterdam"
      
      - entity_id: "e3"
        text: "info@rijksmuseum.nl"
        entity_type: "TXT.IEM"
        start_offset: 74
        end_offset: 93
        confidence: 0.99
        ontology_class: "schema:email"
        nerd_class: null
        
        # ENTITY PROVENANCE (5 components) - email from mailto link
        provenance:
          namespace: "schema:"
          xpath: "/html[1]/body[1]/footer[1]/a[@href='mailto:info@rijksmuseum.nl']"
          xpath_match_score: 1.0
          extraction_method: "structural"  # Tier 1 - from HTML attribute
        tier: 1
      
      - entity_id: "e4"
        text: "Dr. Taco Dibbits"
        entity_type: "BEING.STF"
        start_offset: 105
        end_offset: 121
        confidence: 0.92
        ontology_class: "picom:PersonObservation"
        nerd_class: "nerd:Person"
        
        # ENTITY PROVENANCE (5 components)
        provenance:
          namespace: "pico:"
          xpath: "/html[1]/body[1]/main[1]/section[3]/div[1]/h3[1]"
          xpath_match_score: 0.92
          xpath_matched_text: "Dr. Taco Dibbits, General Director"
        
        entity_link:
          reference_uri: "http://www.wikidata.org/entity/Q56314921"
          source: "Wikidata"
          linking_confidence: 0.88
        
        attributes:
          role: "General Director"
          pnv_name:
            honorificPrefix: "Dr."
            givenName: "Taco"
            baseSurname: "Dibbits"
        
        # PROPERTY CLAIMS for staff relationships
        property_claims:
          - claim_id: "urn:glam:claim:e4-role"
            claim_type: "staff_role"
            claim_value: "General Director"
            provenance:
              namespace: "pico:"
              xpath: "/html[1]/body[1]/main[1]/section[3]/div[1]/h3[1]"
              retrieved_on: "2025-12-02T10:30:00Z"
              extraction_agent: "glm4.6"
              extraction_convention: "entity_annotation_rules_v1.6.3"
              html_file: "web/0001/rijksmuseum.nl/en/about-us.html"
              xpath_match_score: 0.75
              xpath_matched_text: "Dr. Taco Dibbits, General Director"
            confidence: 0.525  # 0.75 * 0.7 (NLP weight)
            tier: 3
        
        relationships:
          - predicate: "org:memberOf"
            target_entity_id: "e1"
  
  # COMPLETE PROPERTY CLAIM EXAMPLE (standalone)
  # Shows full structure for property extraction output
  property_claim_example:
    description: |
      Complete example of a property claim with all 5 provenance components.
      This structure is used for property extraction (Section 8) outputs.      
    
    claim:
      claim_id: "urn:glam:claim:550e8400-e29b-41d4-a716-446655440000"
      claim_type: "founding_date"
      claim_value: "1885"
      normalized_value: "1885-01-01"
      
      subject_entity_id: "urn:glam:entity:rijksmuseum"
      
      # === 5 PROVENANCE COMPONENTS (ALL REQUIRED) ===
      #
      # 1. NAMESPACE - ontology prefix for claim predicate
      namespace: "rico:"
      
      # 2. PATH - location in source document
      xpath: "/html[1]/body[1]/main[1]/article[1]/div[2]/p[3]"
      
      # 3. TIMESTAMP - when source was retrieved
      retrieved_on: "2025-12-02T10:30:00Z"
      
      # 4. AGENT - extraction model/script/human
      extraction_agent: "glm4.6"
      
      # 5. CONTEXT CONVENTION - version of extraction rules
      extraction_convention: "entity_annotation_rules_v1.6.3"
      # ==============================================
      
      # Additional provenance fields
      source_url: "https://www.rijksmuseum.nl/nl/over-ons/geschiedenis"
      html_file: "web/0001/rijksmuseum.nl/over-ons/geschiedenis.html"
      
      # Verification (computed, not estimated)
      xpath_match_score: 0.92
      xpath_matched_text: "Het Rijksmuseum is in 1885 opgericht"
      
      # Computed confidence and tier
      confidence: 0.644  # xpath_match_score * method_weight (0.92 * 0.7)
      tier: 3  # NLP extraction

# =============================================================================
# SECTION 12: VALIDATION AND QUALITY
# =============================================================================
# Quality control rules for extraction validation.

validation:

  required_fields:
    - "text"
    - "entity_type"
    - "confidence"
  
  confidence_thresholds:
    high: 0.90
    medium: 0.70
    low: 0.50
    minimum: 0.30
  
  rules:
    - id: "VAL001"
      rule: "Entity text must not be empty"
    
    - id: "VAL002"
      rule: "Entity type must be a valid type code from this convention"
    
    - id: "VAL003"
      rule: "Confidence must be between 0.0 and 1.0"
    
    - id: "VAL004"
      rule: "Relationships must reference valid entity_ids in same extraction"
    
    - id: "VAL005"
      rule: "HERINST entities should have at least one contact relationship"
      applies_to: "ORG.HER"
      recommendation: "Link to address, email, website, or social media"
    
    - id: "VAL006"
      rule: "Staff entities should link to their institution"
      applies_to: "BEING.STF"
      required_relationship: "org:memberOf to ORG.HER"

# =============================================================================
# SECTION 13: PROVENANCE TRACKING
# =============================================================================
# Provenance metadata for extraction audit trail.

provenance:

  description: |
    Every extraction must include provenance metadata documenting the source,
    method, and processing chain. This enables quality assessment and
    reproducibility.    

  required_metadata:
    source_url:
      description: "URL of the source document/page"
      type: "string"
      format: "uri"
    
    extraction_timestamp:
      description: "ISO 8601 timestamp of extraction"
      type: "string"
      format: "date-time"
    
    convention_version:
      description: "Version of this NER convention used"
      type: "string"
      value: "1.6.3-unified"
    
    extractor_id:
      description: "Identifier for the extraction system/model"
      type: "string"
      examples:
        - "claude-3.5-sonnet"
        - "glam-ner-v1"
    
    html_archive_path:
      description: "Path to archived HTML source (for XPath verification)"
      type: "string"

  xpath_provenance:
    description: |
      For web extractions, each entity should include XPath to its location
      in the archived HTML. This enables verification and quality control.      
    
    fields:
      xpath:
        description: "XPath expression locating the entity in HTML"
        type: "string"
      xpath_match_score:
        description: "1.0 for exact match, less than 1.0 for fuzzy"
        type: "number"

# =============================================================================
# SECTION 14: FRICTION POINTS AND EDGE CASES
# =============================================================================
# Known difficult cases and resolution strategies.

friction_points:

  - id: "FP001"
    title: "Organization vs. Building ambiguity"
    description: |
      Heritage institutions are both organizations and physical locations.
      The Rijksmuseum can refer to the institution or the building.      
    resolution: |
      Default to ORG.HER for heritage institutions. Apply double-tagging
      (ORG.HER + PLC.BLD) only when the physical building is specifically
      being discussed (e.g., the Rijksmuseum building was renovated).      

  - id: "FP002"
    title: "Historical vs. current names"
    description: |
      Institutions may be referenced by historical names that differ from
      current official names.      
    resolution: |
      Tag the name as it appears in the text. Use the same entity type
      regardless of whether historical or current. Note name changes in
      relationship metadata if detectable.      

  - id: "FP003"
    title: "Abbreviated institution names"
    description: |
      Common abbreviations like the Rijks or MoMA may not be explicit.      
    resolution: |
      Tag abbreviations as ORG.HER when context makes the institution clear.
      Lower confidence score if abbreviation is ambiguous.      

  - id: "FP004"
    title: "Generic role vs. named staff"
    description: |
      Distinguishing between the director (generic) and Director Jan de Wit
      (specific person in role).      
    resolution: |
      Tag only when a proper name is present. The director alone is excluded
      per BEI_EXC001. Director Jan de Wit is tagged as BEING.STF.      

  - id: "FP005"
    title: "Nested locations"
    description: |
      Addresses contain multiple location levels: street, city, country.      
    resolution: |
      Tag the complete address as a single PLC.IAD or PLC.ADR entity.
      Do NOT separately tag city or country within an address span.
      Tag cities/countries separately only when they appear independently.      

  - id: "FP006"
    title: "Date formats across cultures"
    description: |
      Dates appear in many formats: 15 July 1606, 07/15/1606, 1606-07-15      
    resolution: |
      Tag complete date expressions as single TMP.DAT entities regardless
      of format. Include the full expression as it appears.      

  - id: "FP007"
    title: "Collection names vs. general descriptions"
    description: |
      Distinguishing the Dutch Masters Collection (proper name) from
      the collection of Dutch masters (description).      
    resolution: |
      Tag capitalized, proper-noun collection names as DEN.COL.
      Do not tag lowercase generic descriptions.      

# =============================================================================
# SECTION 15: DOCUMENT STRUCTURE AND NAMESPACE PATHS
# =============================================================================
# CRITICAL: Namespaces and document paths are ESSENTIAL for clustering entities
# and distinguishing context-specific entity relationships.

document_structure:
  
  purpose: |
    Entity annotations occur within STRUCTURED DOCUMENTS - not flat text streams.
    The LOCATION of an entity within document structure determines:
    
    1. SEMANTIC SCOPE: An entity in a header governs entities in subsequent paragraphs
    2. CO-REFERENCE CLUSTERS: Entities in the same structural unit likely co-refer
    3. RELATIONSHIP CONTEXT: Header-paragraph relations differ from paragraph-paragraph
    4. PROVENANCE PRECISION: XPath/JSONPath enables exact location for verification
    
    Without namespace paths, entity extraction loses critical context that
    distinguishes "the director" in a museum description from "the director"
    in a film credits section of the same document.    
  
  design_principles:
    - principle: "STRUCTURE IS MEANING"
      description: |
        Document structure (headers, sections, paragraphs, lists) carries semantic
        information. A person mentioned in a "Board of Directors" section has a
        different relationship to the organization than a person mentioned in a
        "Historical Overview" section.        
    
    - principle: "PATHS ENABLE CLUSTERING"
      description: |
        Entities sharing a common path prefix belong to the same structural context.
        Clustering by path enables:
        - Scoped entity resolution (disambiguate within section before document)
        - Contextual relationship inference (section membership implies relationship)
        - Provenance aggregation (all claims from same region share reliability)        
    
    - principle: "NAMESPACES PREVENT COLLISION"
      description: |
        The same entity mention in different structural contexts may require
        different annotations or link to different knowledge base entities.
        Namespaces ensure annotations are addressable without ambiguity.        
    
    - principle: "LAYOUT INFORMS SEMANTICS"
      description: |
        Visual layout (sidebars, captions, footnotes, marginalia) carries meaning
        distinct from main body text. PAGE-XML text regions, HTML semantic elements,
        and JSON structural keys all encode layout that affects interpretation.        
  
  # ---------------------------------------------------------------------------
  # LAYOUT SEMANTIC ONTOLOGY
  # ---------------------------------------------------------------------------
  # Comprehensive hypernym classes and properties for document structure annotation.
  # These classes are format-agnostic and apply to PAGE-XML, HTML, JSON, MD, EPUB, PDF.
  # 
  # PRIMARY AUTHORITIES:
  # - W3C Web Annotation Data Model (oa:) - annotation targeting and selectors
  # - CIDOC-CRM (crm:) - information objects and carriers
  # - RiC-O (rico:) - record parts and structure
  # - PREMIS (premis:) - digital object hierarchy
  # - Dublin Core (dcterms:) - part/whole relationships
  # - NIF 2.0 (nif:) - text annotation interchange
  # ---------------------------------------------------------------------------
  
  layout_semantic_ontology:
    
    purpose: |
      This ontology defines semantic classes for document layout elements that
      generative AI models can use to annotate/describe structural context.
      
      CRITICAL: Layout claims are DISTINCT from entity claims. A complete annotation
      requires TWO NESTED LAYERS of provenance:
      
      1. LAYOUT CLAIM: "This text region is a sidebar" 
         - Has its own provenance (XPath, timestamp, confidence, agent)
         - May be uncertain (is this a sidebar or a marginalia?)
      
      2. ENTITY CLAIM: "This sidebar contains person name 'Rembrandt'"
         - Has its own provenance (span offsets, NER model, confidence)
         - REFERENCES the layout claim as context
      
      This separation enables:
      - Independent validation of layout vs. entity extraction
      - Different confidence levels for structure vs. content
      - Reasoning about how layout affects entity interpretation      
    
    # -------------------------------------------------------------------------
    # HYPERNYM: DOCUMENT_REGION (DOC)
    # -------------------------------------------------------------------------
    # The top-level class for any identifiable region within a document.
    # Analogous to CIDOC-CRM E73 Information Object but focused on structure.
    # -------------------------------------------------------------------------
    
    DOCUMENT_REGION:
      code: "DOC"
      definition: |
        A discrete, identifiable region within a document that contains content
        and has structural relationships to other regions. Document regions are
        the fundamental unit of layout annotation.
        
        Every document region has:
        - BOUNDARIES: Start/end positions (character offsets, coordinates, paths)
        - CONTAINMENT: Parent regions that contain it, child regions it contains
        - SEQUENCE: Position relative to sibling regions
        - SEMANTIC ROLE: Function within the document (header, body, supplement)        
      
      ontology_mappings:
        primary_class: "crm:E73_Information_Object"
        primary_class_definition: |
          CIDOC-CRM E73 Information Object: "This class comprises identifiable 
          immaterial items, such as poems, jokes, data sets, images, texts, 
          multimedia objects, procedural prescriptions, computer program code, 
          algorithm or mathematical formulae, that have an objectively 
          recognizable structure."          
        alternative_classes:
          - class: "rico:RecordPart"
            note: "RiC-O class for component parts of archival records"
          - class: "premis:Bitstream"
            note: "PREMIS class for meaningful data segment within file"
          - class: "oa:TextualBody"
            note: "Web Annotation class for textual content"
        linkml_mapping:
          class_uri: "glam:DocumentRegion"
          exact_mappings:
            - "crm:E73_Information_Object"
          close_mappings:
            - "rico:RecordPart"
            - "premis:Bitstream"
          related_mappings:
            - "oa:TextualBody"
            - "nif:Context"
      
      properties:
        - property: "hasParentRegion"
          uri: "glam:hasParentRegion"
          range: "DocumentRegion"
          owl_mapping: "dcterms:isPartOf"
          description: "The containing region (section contains paragraph)"
        
        - property: "hasChildRegion"
          uri: "glam:hasChildRegion"
          range: "DocumentRegion"
          owl_mapping: "dcterms:hasPart"
          description: "Contained regions (paragraph contains sentences)"
        
        - property: "hasNextSibling"
          uri: "glam:hasNextSibling"
          range: "DocumentRegion"
          owl_mapping: "rico:isOrWasAdjacentTo"
          description: "Next region in document order"
        
        - property: "hasPreviousSibling"
          uri: "glam:hasPreviousSibling"
          range: "DocumentRegion"
          description: "Previous region in document order"
        
        - property: "hasSemanticRole"
          uri: "glam:hasSemanticRole"
          range: "LayoutSemanticRole"
          description: "The semantic function of this region"
        
        - property: "regionPath"
          uri: "glam:regionPath"
          range: "xsd:string"
          owl_mapping: "oa:hasSelector"
          description: "XPath, JSONPath, or other path expression"
        
        - property: "regionStart"
          uri: "glam:regionStart"
          range: "xsd:integer"
          owl_mapping: "nif:beginIndex"
          description: "Character offset of region start"
        
        - property: "regionEnd"
          uri: "glam:regionEnd"
          range: "xsd:integer"
          owl_mapping: "nif:endIndex"
          description: "Character offset of region end"
      
      subcategories:
        
        # ---------------------------------------------------------------------
        # PRIMARY CONTENT REGIONS
        # ---------------------------------------------------------------------
        
        DOC.HDR:
          name: "HEADER"
          definition: |
            Heading or title region that introduces and governs subsequent content.
            Headers establish scope for entity interpretation and relationship inference.
            
            Includes: h1-h6, chapter titles, section headings, running headers,
            PAGE-XML TextRegion[@type='header'], JSON object keys.            
          
          ontology_mappings:
            primary_class: "schema:headline"
            alternative_classes:
              - "bf:Title"
              - "dcterms:title"
            nif_mapping: "nif:Title"
          
          header_levels:
            - level: 1
              scope: "document"
              html: "h1"
              pagexml: "TextRegion[@type='heading'][@level='1']"
              semantic: "Document title or primary topic"
            - level: 2
              scope: "chapter"
              html: "h2"
              pagexml: "TextRegion[@type='heading'][@level='2']"
              semantic: "Major section or chapter heading"
            - level: 3
              scope: "section"
              html: "h3"
              semantic: "Subsection heading"
            - level: 4
              scope: "subsection"
              html: "h4"
              semantic: "Sub-subsection heading"
            - level: 5
              scope: "paragraph_group"
              html: "h5"
              semantic: "Minor heading or label"
            - level: 6
              scope: "inline"
              html: "h6"
              semantic: "Inline heading or run-in head"
          
          governing_properties:
            - property: "governs"
              uri: "glam:governs"
              range: "DocumentRegion"
              description: "Regions semantically governed by this header"
            - property: "governsUntil"
              uri: "glam:governsUntil"
              range: "DocumentRegion"
              description: "Region where governance ends (next same-level header)"
        
        DOC.PAR:
          name: "PARAGRAPH"
          definition: |
            Block of continuous prose text forming a logical unit of discourse.
            The primary content-bearing unit in most documents.
            
            Includes: HTML <p>, PAGE-XML TextRegion[@type='paragraph'],
            text separated by blank lines, JSON string values.            
          
          ontology_mappings:
            primary_class: "crm:E33_Linguistic_Object"
            alternative_classes:
              - "schema:Text"
              - "nif:Paragraph"
          
          paragraph_properties:
            - property: "paragraphIndex"
              uri: "glam:paragraphIndex"
              range: "xsd:integer"
              description: "Zero-based index within containing section"
            - property: "sentenceCount"
              uri: "glam:sentenceCount"
              range: "xsd:integer"
              description: "Number of sentences in paragraph"
        
        DOC.SEN:
          name: "SENTENCE"
          definition: |
            A grammatical sentence within a paragraph. The minimal unit for
            syntactic analysis and relationship extraction.            
          
          ontology_mappings:
            primary_class: "nif:Sentence"
            alternative_classes:
              - "crm:E33_Linguistic_Object"
          
          sentence_properties:
            - property: "sentenceIndex"
              uri: "glam:sentenceIndex"
              range: "xsd:integer"
              description: "Zero-based index within containing paragraph"
        
        DOC.LST:
          name: "LIST"
          definition: |
            Ordered or unordered enumeration of items sharing parallel structure.
            
            Includes: HTML <ul>/<ol>, bulleted/numbered lists, PAGE-XML lists,
            JSON arrays, markdown lists.            
          
          ontology_mappings:
            primary_class: "schema:ItemList"
            alternative_classes:
              - "rdf:List"
              - "rdf:Seq"
          
          list_types:
            - type: "ordered"
              html: "ol"
              semantic: "Sequence with meaningful order"
            - type: "unordered"
              html: "ul"
              semantic: "Set without meaningful order"
            - type: "definition"
              html: "dl"
              semantic: "Term-definition pairs"
        
        DOC.LIT:
          name: "LIST_ITEM"
          definition: |
            Individual item within a list. Shares type/relationship with siblings.            
          
          ontology_mappings:
            primary_class: "schema:ListItem"
          
          list_item_properties:
            - property: "itemIndex"
              uri: "glam:itemIndex"
              range: "xsd:integer"
              description: "Position within list (0-based)"
            - property: "itemLabel"
              uri: "glam:itemLabel"
              range: "xsd:string"
              description: "Bullet, number, or marker text"
        
        DOC.TBL:
          name: "TABLE"
          definition: |
            Structured tabular data with rows and columns.
            
            Includes: HTML <table>, PAGE-XML tables, markdown tables,
            CSV/TSV content, JSON arrays of objects.            
          
          ontology_mappings:
            primary_class: "schema:Table"
            alternative_classes:
              - "csvw:Table"
          
          table_structure:
            - component: "TABLE_HEADER"
              code: "DOC.TBL.HDR"
              html: "thead, th"
              description: "Column headers defining semantics"
            - component: "TABLE_BODY"
              code: "DOC.TBL.BDY"
              html: "tbody"
              description: "Data rows"
            - component: "TABLE_ROW"
              code: "DOC.TBL.ROW"
              html: "tr"
              description: "Single row of cells"
            - component: "TABLE_CELL"
              code: "DOC.TBL.CEL"
              html: "td"
              description: "Individual data cell"
        
        # ---------------------------------------------------------------------
        # SUPPLEMENTARY CONTENT REGIONS
        # ---------------------------------------------------------------------
        
        DOC.SDB:
          name: "SIDEBAR"
          definition: |
            Supplementary content presented alongside main text, typically
            providing context, metadata, or related information.
            
            HYPERNYM for: marginalia, infoboxes, callout boxes, pull quotes,
            asides, floating content.
            
            Includes: HTML <aside>, Wikipedia infoboxes, PAGE-XML marginal regions,
            floating boxes, pull quotes.            
          
          ontology_mappings:
            primary_class: "schema:WPSideBar"
            alternative_classes:
              - "html:aside"
              - "crm:E73_Information_Object"
          
          hyponyms:
            
            DOC.SDB.MRG:
              name: "MARGINALIA"
              definition: |
                Annotations or notes in the margin of a page. Common in historical
                manuscripts, early printed books, and academic texts.
                
                Marginalia often contain:
                - Reader annotations (comments, corrections)
                - Editorial marks (printer instructions)
                - Cross-references (citations, page numbers)
                - Dates or signatures                
              
              ontology_mappings:
                primary_class: "crm:E34_Inscription"
                note: "Marginalia as inscribed marks on document"
              
              margin_positions:
                - position: "left"
                  pagexml: "TextRegion[@type='marginalia-left']"
                - position: "right"
                  pagexml: "TextRegion[@type='marginalia-right']"
                - position: "top"
                  pagexml: "TextRegion[@type='marginalia-top']"
                - position: "bottom"
                  pagexml: "TextRegion[@type='marginalia-bottom']"
                - position: "interlinear"
                  description: "Between lines of main text"
            
            DOC.SDB.IBX:
              name: "INFOBOX"
              definition: |
                Structured summary box containing key facts about a topic.
                Common in encyclopedias, Wikipedia, and reference works.
                
                Infoboxes contain structured claims with high extraction value.                
              
              ontology_mappings:
                primary_class: "schema:Table"
                note: "Infoboxes are typically key-value tables"
              
              infobox_properties:
                - property: "infoboxType"
                  range: "xsd:string"
                  description: "Type of infobox (person, place, organization)"
            
            DOC.SDB.CLT:
              name: "CALLOUT"
              definition: |
                Highlighted text box drawing attention to key information.
                Includes pull quotes, highlighted passages, tip boxes.                
              
              callout_types:
                - type: "pull_quote"
                  description: "Excerpt from main text displayed prominently"
                - type: "tip"
                  description: "Advice or recommendation"
                - type: "warning"
                  description: "Caution or alert"
                - type: "note"
                  description: "Additional information"
        
        DOC.CAP:
          name: "CAPTION"
          definition: |
            Text describing or explaining a figure, table, or other visual element.
            Captions establish aboutness relationships with visual content.
            
            Includes: HTML <figcaption>, PAGE-XML caption regions, alt text,
            image descriptions.            
          
          ontology_mappings:
            primary_class: "schema:caption"
            alternative_classes:
              - "dcterms:description"
              - "crm:P3_has_note"
          
          caption_properties:
            - property: "captionOf"
              uri: "glam:captionOf"
              range: "DocumentRegion"
              owl_mapping: "schema:about"
              description: "The figure/table this caption describes"
        
        DOC.FTN:
          name: "FOOTNOTE"
          definition: |
            Reference note at bottom of page or end of section/document.
            Contains supplementary information, citations, or clarifications.
            
            Includes: Footnotes, endnotes, sidenotes, margin notes with references.            
          
          ontology_mappings:
            primary_class: "bibo:Note"
            alternative_classes:
              - "schema:Comment"
              - "crm:E33_Linguistic_Object"
          
          footnote_types:
            - type: "footnote"
              location: "bottom_of_page"
              pagexml: "TextRegion[@type='footnote']"
            - type: "endnote"
              location: "end_of_chapter"
            - type: "sidenote"
              location: "margin"
          
          footnote_properties:
            - property: "footnoteMarker"
              uri: "glam:footnoteMarker"
              range: "xsd:string"
              description: "The marker symbol (*, 1, a, etc.)"
            - property: "referencesLocation"
              uri: "glam:referencesLocation"
              range: "DocumentRegion"
              description: "Location of footnote marker in main text"
        
        DOC.FIG:
          name: "FIGURE"
          definition: |
            Single visual content item (image, diagram, chart).
            For collections of images, use DOC.GAL (Gallery).
            For geographic/spatial visualizations, use DOC.MAP (Map).
            
            Entities mentioned in figures require visual analysis (OCR, object detection).
            
            Includes: HTML <figure>, embedded images, diagrams, charts, illustrations.            
          
          ontology_mappings:
            primary_class: "schema:ImageObject"
            alternative_classes:
              - "crm:E38_Image"
              - "edm:WebResource"
              - "foaf:Image"
            linkml_mapping:
              class_uri: "schema:ImageObject"
              exact_mappings:
                - "crm:E38_Image"
              close_mappings:
                - "edm:WebResource"
          
          figure_types:
            - type: "photograph"
              description: "Photographic image"
            - type: "illustration"
              description: "Drawing, painting, or artwork"
            - type: "diagram"
              description: "Schematic, flowchart, or technical drawing"
            - type: "chart"
              description: "Data visualization (bar, line, pie, etc.)"
            - type: "scan"
              description: "Digitized physical document or object"
        
        # ---------------------------------------------------------------------
        # MEDIA COLLECTION REGIONS
        # ---------------------------------------------------------------------
        
        DOC.GAL:
          name: "GALLERY"
          code: "DOC.GAL"
          definition: |
            Collection of related visual or media items presented as a unit.
            Distinguished from single figures by containing MULTIPLE items with
            shared context, navigation, or thematic grouping.
            
            Domain-agnostic applications:
            - Heritage: Manuscript illuminations, artifact photo sets, exhibition views
            - Web: Image carousels, product galleries, portfolio showcases
            - Publishing: Photo essays, plate sections, illustration series
            - Archives: Document set scans, correspondence series images
            - Museums: Object photography series, conservation documentation
            
            Entities in galleries share contextual scope from gallery title/caption.
            Individual items may have their own DOC.FIG claims nested within.            
          
          ontology_mappings:
            primary_class: "schema:ImageGallery"
            alternative_classes:
              - "schema:Collection"
              - "crm:E78_Curated_Holding"
              - "edm:Aggregation"
              - "as:Collection"  # ActivityStreams
            linkml_mapping:
              class_uri: "schema:ImageGallery"
              close_mappings:
                - "schema:Collection"
                - "crm:E78_Curated_Holding"
              related_mappings:
                - "edm:Aggregation"
          
          format_mappings:
            html:
              elements: ["div[class*='gallery']", "div[class*='carousel']", "div[class*='slider']", "ul[class*='gallery']"]
              aria_roles: ["group", "listbox"]
            json:
              patterns: ["$.gallery", "$.images[]", "$.media.items[]"]
            pagexml:
              note: "Rare in PAGE-XML; group of consecutive ImageRegions"
          
          gallery_types:
            - type: "image_gallery"
              description: "Collection of photographs or illustrations"
              examples: ["museum object photos", "manuscript folios", "exhibition views"]
            - type: "carousel"
              description: "Horizontally/vertically scrolling media set"
            - type: "slideshow"
              description: "Sequential presentation with transitions"
            - type: "lightbox"
              description: "Thumbnail grid with modal expansion"
            - type: "filmstrip"
              description: "Linear sequence of video thumbnails or stills"
            - type: "plate_section"
              description: "Bound illustration pages in printed works"
              note: "Common in historical scientific/art publications"
          
          gallery_properties:
            - property: "galleryTitle"
              uri: "schema:name"
              description: "Title of the gallery collection"
            - property: "itemCount"
              uri: "schema:numberOfItems"
              description: "Number of items in gallery"
            - property: "galleryItems"
              uri: "schema:hasPart"
              range: "DocumentRegion"
              description: "Individual items (DOC.FIG) within gallery"
            - property: "curatedBy"
              uri: "schema:curator"
              owl_mapping: "prov:wasAttributedTo"
              description: "Agent who assembled the gallery"
          
          entity_extraction_notes: |
            Galleries present special extraction challenges:
            1. Gallery TITLE provides context for all contained items
            2. Individual CAPTIONS may override gallery-level context
            3. Items may be ORDERED (narrative sequence) or UNORDERED (thematic group)
            4. Navigation elements (prev/next, thumbnails) are structural, not content
            
            Recommended approach:
            - Create layout claim for gallery container (DOC.GAL)
            - Create nested layout claims for each item (DOC.FIG)
            - Entity claims reference their immediate container
            - Inherit gallery context when item lacks own caption            
        
        DOC.MAP:
          name: "MAP"
          code: "DOC.MAP"
          definition: |
            Cartographic or spatial representation of geographic information.
            Distinguished from generic figures by explicit spatial semantics.
            
            CRITICAL DISTINCTION from GEOMETRY (GEO) entity type:
            - DOC.MAP is a DOCUMENT REGION containing a map visualization
            - GEO is an ENTITY TYPE for coordinate/shape data extracted from any region
            
            Domain-agnostic applications:
            - Heritage: Historical maps, archaeological site plans, building floorplans
            - Web: Interactive maps, location widgets, route displays
            - Publishing: Atlas plates, thematic maps, navigation charts
            - Archives: Survey maps, cadastral records, military charts
            - Urban planning: Zoning maps, infrastructure layouts
            
            Entities in maps require spatial reasoning for extraction.
            Place names (TOP) may appear as labels; geometry (GEO) as shapes/points.            
          
          ontology_mappings:
            primary_class: "crm:E73_Information_Object"
            alternative_classes:
              - "schema:Map"
              - "edm:WebResource"
              - "geosparql:SpatialObject"
              - "bibo:Map"
            linkml_mapping:
              class_uri: "schema:Map"
              exact_mappings:
                - "bibo:Map"
              close_mappings:
                - "crm:E73_Information_Object"
              related_mappings:
                - "geosparql:SpatialObject"
          
          format_mappings:
            html:
              elements: ["div[class*='map']", "div[id*='map']", "leaflet-container", "mapboxgl-map", "gm-style"]
              note: "Interactive maps often use Leaflet, Mapbox, Google Maps containers"
            json:
              patterns: ["$.map", "$.geojson", "$.geometry", "$.features[]"]
              standards: ["GeoJSON (RFC 7946)", "TopoJSON"]
            image_formats:
              extensions: [".geotiff", ".tiff", ".jpg", ".png"]
              note: "Georeferenced images contain embedded coordinate metadata"
            pagexml:
              type: "GraphicRegion[@type='map']"
          
          map_types:
            - type: "reference_map"
              description: "General purpose geographic reference"
              examples: ["world map", "country outline", "street map"]
            - type: "thematic_map"
              description: "Data visualization on geographic base"
              examples: ["choropleth", "heat map", "dot distribution"]
            - type: "historical_map"
              description: "Map from historical period (primary source)"
              examples: ["17th century nautical chart", "medieval mappa mundi"]
            - type: "site_plan"
              description: "Architectural or archaeological layout"
              examples: ["floor plan", "excavation grid", "campus map"]
            - type: "route_map"
              description: "Navigation or journey visualization"
              examples: ["transit map", "pilgrimage route", "trade route"]
            - type: "cadastral_map"
              description: "Property boundary documentation"
              examples: ["land survey", "parcel map", "deed plat"]
            - type: "interactive_map"
              description: "User-manipulable web map with layers/zoom"
              examples: ["Leaflet widget", "Google Maps embed", "IIIF geo extension"]
          
          map_components:
            - component: "MAP.BAS"
              name: "Base Layer"
              description: "Background geographic reference (satellite, street, terrain)"
            - component: "MAP.OVL"
              name: "Overlay Layer"
              description: "Thematic data layer on top of base"
            - component: "MAP.MRK"
              name: "Markers"
              description: "Point features (pins, icons, labels)"
            - component: "MAP.SHP"
              name: "Shapes"
              description: "Polygon/polyline features (boundaries, routes)"
            - component: "MAP.LEG"
              name: "Legend"
              description: "Symbol key and scale information"
            - component: "MAP.CTL"
              name: "Controls"
              description: "Zoom, pan, layer toggles (structural, not content)"
          
          map_properties:
            - property: "mapTitle"
              uri: "schema:name"
              description: "Title of the map"
            - property: "spatialCoverage"
              uri: "schema:spatialCoverage"
              description: "Geographic extent represented"
            - property: "mapScale"
              uri: "schema:contentSize"
              description: "Representative fraction or verbal scale"
            - property: "projection"
              uri: "geosparql:hasSpatialAccuracy"
              description: "Cartographic projection used"
            - property: "temporalCoverage"
              uri: "schema:temporalCoverage"
              description: "Time period depicted (for historical maps)"
          
          entity_extraction_notes: |
            Maps present unique extraction challenges:
            
            1. TOPONYMS (TOP) appear as:
               - Labels on map face
               - Legend entries
               - Title/caption text
               - Popup/tooltip content (interactive maps)
            
            2. GEOMETRY (GEO) data includes:
               - Point coordinates (markers)
               - Bounding boxes
               - Polygon vertices (regions, buildings)
               - Polylines (routes, rivers, borders)
            
            3. TEMPORAL context:
               - Historical maps show past geography (not current!)
               - Map creation date ≠ depicted time period
               - Boundary changes over time
            
            4. PROVENANCE considerations:
               - Georeferenced scans have transformation accuracy
               - Interactive maps have tile source attribution
               - Derived maps inherit source map provenance
            
            Recommended approach:
            - Create layout claim for map container (DOC.MAP)
            - Extract toponyms with map-specific confidence (label legibility)
            - Extract geometry with coordinate reference system metadata
            - Track both map creation date AND depicted time period            
        
        # ---------------------------------------------------------------------
        # AUDIOVISUAL CONTENT REGIONS
        # ---------------------------------------------------------------------
        
        DOC.AUD:
          name: "AUDIO"
          code: "DOC.AUD"
          definition: |
            Audio content region (sound recording, podcast, music, oral history).
            Entities require speech-to-text or audio analysis for extraction.
            
            Domain-agnostic applications:
            - Heritage: Oral history recordings, ethnographic field recordings
            - Web: Podcast episodes, music players, audio articles
            - Archives: Radio broadcasts, interview recordings, speeches
            - Museums: Audio guides, soundscapes, musical instrument recordings
            - Linguistics: Language documentation, dialect samples
            
            Audio content often has associated TRANSCRIPT (DOC.PAR) which is the
            primary source for NER; audio itself provides prosodic/speaker metadata.            
          
          ontology_mappings:
            primary_class: "schema:AudioObject"
            alternative_classes:
              - "crm:E73_Information_Object"
              - "edm:WebResource"
              - "premis:IntellectualEntity"
            linkml_mapping:
              class_uri: "schema:AudioObject"
              close_mappings:
                - "crm:E73_Information_Object"
          
          format_mappings:
            html:
              elements: ["audio", "div[class*='audio-player']", "div[class*='podcast']"]
            json:
              patterns: ["$.audio", "$.podcast", "$.episodes[]", "$.tracks[]"]
            file_formats:
              extensions: [".mp3", ".wav", ".ogg", ".flac", ".m4a", ".aac"]
          
          audio_types:
            - type: "speech"
              description: "Spoken word content"
              examples: ["interview", "lecture", "oral history", "podcast"]
            - type: "music"
              description: "Musical performance or composition"
              examples: ["concert recording", "album track", "folk song"]
            - type: "soundscape"
              description: "Environmental or ambient audio"
              examples: ["field recording", "museum ambiance", "historic sound"]
            - type: "narration"
              description: "Scripted audio guide or documentary"
              examples: ["museum audio guide", "audiobook chapter"]
          
          audio_properties:
            - property: "duration"
              uri: "schema:duration"
              description: "Length of audio content (ISO 8601 duration)"
            - property: "transcript"
              uri: "schema:transcript"
              range: "DocumentRegion"
              description: "Text transcription of audio (DOC.PAR)"
            - property: "speaker"
              uri: "schema:actor"
              description: "Person(s) speaking in recording"
            - property: "recordingDate"
              uri: "schema:dateCreated"
              description: "When audio was recorded"
            - property: "recordingLocation"
              uri: "schema:contentLocation"
              description: "Where audio was recorded"
          
          audio_segments:
            - segment: "AUD.SPK"
              name: "Speaker Turn"
              description: "Contiguous speech by single speaker"
            - segment: "AUD.MUS"
              name: "Music Segment"
              description: "Musical interlude or background"
            - segment: "AUD.SIL"
              name: "Silence"
              description: "Intentional pause or gap"
            - segment: "AUD.SFX"
              name: "Sound Effect"
              description: "Non-speech, non-music audio event"
          
          entity_extraction_notes: |
            Audio requires multi-modal extraction:
            
            1. TRANSCRIPT-BASED (primary):
               - Extract entities from associated transcript (DOC.PAR)
               - Timestamps align text spans to audio segments
               - Speaker diarization links entities to speakers
            
            2. AUDIO-DIRECT (secondary):
               - Named entity recognition from ASR output
               - Speaker identification (voice biometrics)
               - Language/dialect detection
               - Prosodic analysis (emphasis, emotion)
            
            3. METADATA (tertiary):
               - ID3 tags (music files)
               - Episode metadata (podcasts)
               - Catalog records (archives)
            
            Provenance must track:
            - ASR model and confidence
            - Human transcription vs. automated
            - Timestamp precision            
        
        DOC.VID:
          name: "VIDEO"
          code: "DOC.VID"
          definition: |
            Video content region (moving image with or without audio).
            Entities require multimodal analysis: visual, audio, and text (captions/OCR).
            
            Domain-agnostic applications:
            - Heritage: Documentary footage, conservation documentation, exhibition videos
            - Web: Embedded videos, live streams, video articles
            - Archives: News broadcasts, home movies, surveillance footage
            - Museums: Artist interviews, performance recordings, virtual tours
            - Education: Lecture recordings, tutorials, demonstrations
            
            Videos combine visual (DOC.FIG), audio (DOC.AUD), and textual modalities.
            Entities may appear in any modality and require cross-modal linking.            
          
          ontology_mappings:
            primary_class: "schema:VideoObject"
            alternative_classes:
              - "crm:E73_Information_Object"
              - "edm:WebResource"
              - "premis:IntellectualEntity"
            linkml_mapping:
              class_uri: "schema:VideoObject"
              close_mappings:
                - "crm:E73_Information_Object"
          
          format_mappings:
            html:
              elements: ["video", "iframe[src*='youtube']", "iframe[src*='vimeo']", "div[class*='video-player']"]
            json:
              patterns: ["$.video", "$.videos[]", "$.media.video"]
            file_formats:
              extensions: [".mp4", ".webm", ".mov", ".avi", ".mkv"]
          
          video_types:
            - type: "documentary"
              description: "Non-fiction narrative video"
            - type: "interview"
              description: "Recorded conversation with subject"
            - type: "performance"
              description: "Recorded artistic performance"
            - type: "instructional"
              description: "Tutorial or demonstration"
            - type: "surveillance"
              description: "Continuous monitoring footage"
            - type: "news"
              description: "Broadcast journalism content"
            - type: "archival"
              description: "Historical footage or home movies"
          
          video_properties:
            - property: "duration"
              uri: "schema:duration"
              description: "Length of video (ISO 8601 duration)"
            - property: "transcript"
              uri: "schema:transcript"
              range: "DocumentRegion"
              description: "Text transcription of speech"
            - property: "caption"
              uri: "schema:caption"
              range: "DocumentRegion"
              description: "Closed captions or subtitles"
            - property: "thumbnail"
              uri: "schema:thumbnail"
              description: "Representative still image"
          
          video_segments:
            - segment: "VID.SCN"
              name: "Scene"
              description: "Continuous action segment"
            - segment: "VID.SHT"
              name: "Shot"
              description: "Single camera take"
            - segment: "VID.TTL"
              name: "Title Card"
              description: "Text overlay or title screen"
            - segment: "VID.CRD"
              name: "Credits"
              description: "Attribution information"
        
        DOC.EMB:
          name: "EMBEDDED_INTERACTIVE"
          code: "DOC.EMB"
          definition: |
            Embedded interactive content from external source or rich application.
            Includes iframes, widgets, web components, and embedded applications.
            
            Domain-agnostic applications:
            - Heritage: IIIF viewers, 3D model viewers, timeline widgets
            - Web: Social media embeds, code playgrounds, data visualizations
            - Publishing: Interactive charts, explorable explanations
            - Museums: Virtual tours, interactive exhibits, AR/VR content
            - Archives: Document viewers, transcription interfaces
            
            Entities within embedded content may require API access or headless
            browser rendering to extract; simple iframe inspection is insufficient.            
          
          ontology_mappings:
            primary_class: "schema:WebApplication"
            alternative_classes:
              - "schema:CreativeWork"
              - "as:Application"
            linkml_mapping:
              class_uri: "schema:WebApplication"
          
          format_mappings:
            html:
              elements: ["iframe", "embed", "object", "web-component", "[data-widget]"]
            json:
              patterns: ["$.embed", "$.widget", "$.interactive"]
          
          embed_types:
            - type: "iiif_viewer"
              description: "IIIF-compliant image viewer"
              examples: ["Mirador", "Universal Viewer", "OpenSeadragon"]
              note: "Extract manifest URL for entity-rich metadata"
            - type: "3d_viewer"
              description: "3D model visualization"
              examples: ["Sketchfab embed", "Three.js scene", "WebGL viewer"]
            - type: "timeline"
              description: "Temporal data visualization"
              examples: ["TimelineJS", "Chronoline", "d3 timeline"]
            - type: "social_embed"
              description: "Social media post or feed"
              examples: ["Twitter/X embed", "Instagram post", "YouTube video"]
            - type: "data_viz"
              description: "Interactive data visualization"
              examples: ["D3.js chart", "Plotly graph", "Tableau embed"]
            - type: "code_playground"
              description: "Executable code environment"
              examples: ["CodePen", "JSFiddle", "Observable notebook"]
            - type: "form"
              description: "Interactive input form"
              examples: ["Survey", "search widget", "booking form"]
            - type: "virtual_tour"
              description: "360° or spatial navigation experience"
              examples: ["Google Street View", "Matterport", "museum virtual tour"]
          
          embed_properties:
            - property: "embedSource"
              uri: "schema:embedUrl"
              description: "URL of embedded content"
            - property: "embedProvider"
              uri: "schema:provider"
              description: "Service providing the embed"
            - property: "embedType"
              uri: "glam:embedType"
              description: "Category of interactive content"
            - property: "requiresInteraction"
              uri: "glam:requiresInteraction"
              description: "Whether user action needed to reveal content"
          
          entity_extraction_notes: |
            Embedded content presents significant extraction challenges:
            
            1. ACCESSIBILITY:
               - Content may be behind authentication
               - May require JavaScript execution
               - May load asynchronously
               - Cross-origin policies may block access
            
            2. STRATEGIES by embed type:
               - IIIF: Fetch manifest JSON for rich metadata
               - Social: Use platform APIs (Twitter, Instagram)
               - Data viz: Extract from data source if accessible
               - 3D: Parse model metadata, texture labels
            
            3. PROVENANCE requirements:
               - Record embed source URL
               - Record extraction timestamp (content may change)
               - Record access method (API, headless browser, etc.)
               - Note if content was inaccessible
            
            4. FALLBACK strategies:
               - Use surrounding context (embed caption, link text)
               - Use embed URL structure for hints
               - Use embed provider metadata            
        
        # ---------------------------------------------------------------------
        # NAVIGATION AND STRUCTURE REGIONS
        # ---------------------------------------------------------------------
        
        DOC.NAV:
          name: "NAVIGATION"
          definition: |
            Generic navigational elements (breadcrumbs, menus, links).
            For specific navigation structures, use dedicated types:
            - DOC.TOC for tables of contents
            - DOC.IDX for indices            
          
          ontology_mappings:
            primary_class: "schema:SiteNavigationElement"
            linkml_mapping:
              class_uri: "schema:SiteNavigationElement"
          
          navigation_types:
            - type: "breadcrumb"
              name: "Breadcrumb Trail"
              html: "nav[aria-label='breadcrumb'], .breadcrumb"
            - type: "menu"
              name: "Navigation Menu"
              html: "nav, menu"
            - type: "sitemap"
              name: "Site Map"
              description: "Hierarchical site structure overview"
        
        DOC.TOC:
          name: "TABLE_OF_CONTENTS"
          code: "DOC.TOC"
          definition: |
            Structured listing of document sections with page/location references.
            Distinguished from general navigation by its document-internal scope
            and hierarchical structure reflecting document organization.
            
            Domain-agnostic applications:
            - Heritage: Manuscript tables, register indices, finding aid outlines
            - Publishing: Book TOCs, journal issue contents, report outlines
            - Web: Article outlines, documentation navigation, wiki page contents
            - Archives: Folder listings, series descriptions, container lists
            - Legal: Statute tables, case indices, contract section lists
            
            High-value for entity extraction: section titles often contain
            key entities (names, places, dates, topics) that scope content.            
          
          ontology_mappings:
            primary_class: "bibo:DocumentPart"
            alternative_classes:
              - "schema:ItemList"
              - "crm:E73_Information_Object"
            linkml_mapping:
              class_uri: "bibo:DocumentPart"
              close_mappings:
                - "schema:ItemList"
          
          format_mappings:
            html:
              elements: ["nav[role='doc-toc']", ".toc", "#table-of-contents", "ol.toc"]
            pagexml:
              type: "TextRegion[@type='table-of-contents']"
            tei:
              element: "<divGen type='toc'/>"
            epub:
              element: "nav[epub:type='toc']"
          
          toc_types:
            - type: "main_toc"
              description: "Primary document table of contents"
            - type: "list_of_figures"
              description: "Figure/illustration listing with page numbers"
              alias: "LOF"
            - type: "list_of_tables"
              description: "Table listing with page numbers"
              alias: "LOT"
            - type: "list_of_abbreviations"
              description: "Abbreviation/acronym listing"
            - type: "list_of_maps"
              description: "Map listing with page numbers"
            - type: "list_of_plates"
              description: "Plate/illustration listing (historical)"
          
          toc_properties:
            - property: "tocEntry"
              uri: "schema:itemListElement"
              description: "Individual TOC entry"
            - property: "sectionTitle"
              uri: "schema:name"
              description: "Title of referenced section"
            - property: "pageReference"
              uri: "bibo:pageStart"
              description: "Page number or location reference"
            - property: "nestingLevel"
              uri: "glam:hierarchyLevel"
              description: "Depth in TOC hierarchy (1=chapter, 2=section, etc.)"
          
          entity_extraction_notes: |
            TOC entries are HIGH-VALUE for entity extraction:
            
            1. SECTION TITLES often contain:
               - Person names (biographical sections)
               - Place names (geographic chapters)
               - Date ranges (chronological sections)
               - Organization names (institutional histories)
            
            2. HIERARCHICAL CONTEXT:
               - Parent entries scope child entries
               - "Part I: The Netherlands" → child sections are Dutch-related
            
            3. PAGE REFERENCES enable:
               - Linking entities to page ranges
               - Validating entity locations in document            
        
        DOC.IDX:
          name: "INDEX"
          code: "DOC.IDX"
          definition: |
            Alphabetical or systematic listing of terms, names, or subjects
            with page/location references. Distinguished from TOC by its
            alphabetical organization and term-based (not section-based) structure.
            
            Domain-agnostic applications:
            - Heritage: Name indices, place indices, subject indices
            - Publishing: Back-of-book indices, periodical indices
            - Archives: Finding aid indices, name authority files
            - Legal: Case citation indices, statute indices
            - Academic: Author indices, keyword indices
            
            EXTREMELY high-value for NER: indices are curated entity lists
            with location references, essentially pre-annotated entity data.            
          
          ontology_mappings:
            primary_class: "bibo:DocumentPart"
            alternative_classes:
              - "schema:ItemList"
              - "skos:Collection"
            linkml_mapping:
              class_uri: "bibo:DocumentPart"
              related_mappings:
                - "skos:Collection"
          
          format_mappings:
            html:
              elements: ["div.index", "#index", "section[role='doc-index']"]
            pagexml:
              type: "TextRegion[@type='index']"
            tei:
              element: "<divGen type='index'/>, <index>"
          
          index_types:
            - type: "name_index"
              description: "Personal and corporate name index"
              entity_type: "AGT"
              note: "Pre-annotated AGENT entities"
            - type: "place_index"
              description: "Geographic and place name index"
              entity_type: "TOP"
              note: "Pre-annotated TOPONYM entities"
            - type: "subject_index"
              description: "Topic and subject index"
              entity_type: "mixed"
            - type: "title_index"
              description: "Work and publication title index"
              entity_type: "WRK"
              note: "Pre-annotated WORK entities"
            - type: "chronological_index"
              description: "Date and event index"
              entity_type: "TMP"
              note: "Pre-annotated TEMPORAL entities"
          
          index_properties:
            - property: "indexTerm"
              uri: "skos:prefLabel"
              description: "The indexed term or name"
            - property: "pageReferences"
              uri: "bibo:pages"
              description: "Page number(s) where term appears"
            - property: "seeAlso"
              uri: "skos:related"
              description: "Cross-reference to related terms"
            - property: "subentry"
              uri: "skos:narrower"
              description: "Nested sub-entries under main term"
          
          entity_extraction_notes: |
            Indices are GOLD STANDARD entity sources:
            
            1. PRE-CURATED ENTITIES:
               - Name indices = curated person/org list
               - Place indices = curated gazetteer
               - Subject indices = controlled vocabulary
            
            2. EXTRACTION STRATEGY:
               - Parse index entries as entity mentions
               - Use index type to assign entity class
               - Page references locate entities in text
            
            3. CROSS-REFERENCE VALUE:
               - "See also" links indicate entity relationships
               - Sub-entries indicate hierarchical relationships            
        
        # ---------------------------------------------------------------------
        # FRONT MATTER REGIONS
        # ---------------------------------------------------------------------
        
        DOC.TTP:
          name: "TITLE_PAGE"
          code: "DOC.TTP"
          definition: |
            Primary identifying page of a document containing title, author,
            publisher, date, and other key metadata. High-value structured
            information source.
            
            Domain-agnostic applications:
            - Heritage: Manuscript title pages, incunabula colophons, broadside headers
            - Publishing: Book title pages, journal covers, report covers
            - Archives: Folder titles, series title sheets
            - Legal: Document covers, contract title pages
            - Academic: Thesis title pages, paper headers
            
            Title pages contain STRUCTURED CLAIMS about the document itself.            
          
          ontology_mappings:
            primary_class: "bibo:DocumentPart"
            alternative_classes:
              - "schema:CoverPage"
              - "crm:E73_Information_Object"
            linkml_mapping:
              class_uri: "bibo:DocumentPart"
          
          format_mappings:
            html:
              elements: [".title-page", "#cover", "section.frontmatter"]
            pagexml:
              type: "TextRegion[@type='title-page']"
            tei:
              element: "<titlePage>"
          
          title_page_components:
            - component: "TTP.TTL"
              name: "Title"
              description: "Main document title"
              ontology: "dcterms:title"
            - component: "TTP.STL"
              name: "Subtitle"
              description: "Secondary title"
              ontology: "bibo:shortTitle"
            - component: "TTP.AUT"
              name: "Author"
              description: "Creator attribution"
              ontology: "dcterms:creator"
            - component: "TTP.PUB"
              name: "Publisher"
              description: "Publisher/printer"
              ontology: "dcterms:publisher"
            - component: "TTP.DAT"
              name: "Date"
              description: "Publication date"
              ontology: "dcterms:date"
            - component: "TTP.PLC"
              name: "Place"
              description: "Publication place"
              ontology: "dcterms:spatial"
            - component: "TTP.EDT"
              name: "Edition"
              description: "Edition statement"
              ontology: "bibo:edition"
            - component: "TTP.IMP"
              name: "Imprint"
              description: "Full publication statement"
              ontology: "bibo:Note"
          
          entity_extraction_notes: |
            Title pages are AUTHORITATIVE entity sources:
            
            1. STRUCTURED CLAIMS about document:
               - Author → AGENT entity with creator role
               - Publisher → GROUP entity
               - Date → TEMPORAL entity (publication date)
               - Place → TOPONYM entity (publication place)
            
            2. HIGH CONFIDENCE: Title page claims are intentional,
               not incidental mentions - treat as authoritative.            
        
        DOC.DED:
          name: "DEDICATION"
          code: "DOC.DED"
          definition: |
            Dedicatory text, epigraph, or inscription typically appearing
            in front matter. May honor a person, quote a source, or set
            thematic context for the work.
            
            Domain-agnostic applications:
            - Heritage: Manuscript dedications, donor inscriptions
            - Publishing: Book dedications, memorial pages
            - Archives: Gift acknowledgments, founding documents
            - Academic: Thesis acknowledgments, memorial lectures
            - Monuments: Dedicatory inscriptions, foundation stones            
          
          ontology_mappings:
            primary_class: "bibo:DocumentPart"
            alternative_classes:
              - "schema:Quotation"
              - "crm:E33_Linguistic_Object"
            linkml_mapping:
              class_uri: "bibo:DocumentPart"
          
          format_mappings:
            html:
              elements: [".dedication", ".epigraph", "blockquote.epigraph"]
            tei:
              element: "<div type='dedication'>, <epigraph>"
          
          dedication_types:
            - type: "dedication"
              description: "Work dedicated to a person or group"
              note: "Dedicatee is an AGENT entity"
            - type: "epigraph"
              description: "Quotation setting thematic context"
              note: "May contain WORK reference (source)"
            - type: "inscription"
              description: "Physical inscription text"
              note: "Common in monuments, buildings"
            - type: "acknowledgment"
              description: "Thanks to supporters/contributors"
              note: "Multiple AGENT entities"
          
          dedication_properties:
            - property: "dedicatee"
              uri: "schema:recipient"
              description: "Person/group to whom work is dedicated"
            - property: "quotationSource"
              uri: "schema:isBasedOn"
              description: "Source of epigraph quotation"
        
        DOC.COL:
          name: "COLOPHON"
          code: "DOC.COL"
          definition: |
            Production statement typically at end of document containing
            printing/publication details: printer, date, place, technical
            specifications. Critical for bibliographic identification.
            
            Domain-agnostic applications:
            - Heritage: Manuscript colophons, incunabula printer statements
            - Publishing: Printer's statements, production credits
            - Archives: Processing notes, digitization metadata
            - Legal: Document certification statements
            - Digital: Software version, generation metadata
            
            Colophons contain AUTHORITATIVE production claims.            
          
          ontology_mappings:
            primary_class: "bibo:DocumentPart"
            alternative_classes:
              - "crm:E65_Creation"
              - "prov:Activity"
            linkml_mapping:
              class_uri: "bibo:DocumentPart"
              related_mappings:
                - "crm:E65_Creation"
          
          format_mappings:
            html:
              elements: [".colophon", "#colophon", "section[role='doc-colophon']"]
            pagexml:
              type: "TextRegion[@type='colophon']"
            tei:
              element: "<colophon>"
          
          colophon_components:
            - component: "COL.PRN"
              name: "Printer"
              description: "Printing house or individual printer"
              ontology: "schema:printer"
            - component: "COL.DAT"
              name: "Date"
              description: "Printing/production date"
              ontology: "dcterms:created"
            - component: "COL.PLC"
              name: "Place"
              description: "Place of production"
              ontology: "dcterms:spatial"
            - component: "COL.TYP"
              name: "Typography"
              description: "Font/type information"
            - component: "COL.PAP"
              name: "Paper"
              description: "Paper/material specifications"
            - component: "COL.CPY"
              name: "Copyright"
              description: "Rights statement"
              ontology: "dcterms:rights"
            - component: "COL.EDN"
              name: "Edition"
              description: "Print run, edition number"
              ontology: "bibo:edition"
          
          entity_extraction_notes: |
            Colophons are CRITICAL for heritage identification:
            
            1. PRODUCTION AGENTS:
               - Printer → GROUP or AGENT entity
               - Publisher → GROUP entity
               - Scribe (manuscripts) → AGENT entity
            
            2. PRODUCTION CONTEXT:
               - Date → TEMPORAL entity
               - Place → TOPONYM entity
            
            3. BIBLIOGRAPHIC VALUE:
               - Often only source for incunabula dating
               - Manuscript colophons name scribes            
        
        # ---------------------------------------------------------------------
        # BACK MATTER REGIONS
        # ---------------------------------------------------------------------
        
        DOC.BIB:
          name: "BIBLIOGRAPHY"
          code: "DOC.BIB"
          definition: |
            List of cited or referenced works. Each entry represents a
            WORK entity with structured bibliographic data.
            
            Domain-agnostic applications:
            - Heritage: Manuscript source lists, catalog references
            - Publishing: Book bibliographies, article references
            - Academic: Citation lists, literature reviews
            - Legal: Case citations, statute references
            - Archives: Related materials, provenance sources
            
            Bibliographies are STRUCTURED WORK entity lists.            
          
          ontology_mappings:
            primary_class: "bibo:DocumentPart"
            alternative_classes:
              - "schema:ItemList"
              - "dcterms:BibliographicResource"
            linkml_mapping:
              class_uri: "bibo:DocumentPart"
          
          format_mappings:
            html:
              elements: [".bibliography", ".references", "#refs", "section[role='doc-bibliography']"]
            tei:
              element: "<listBibl>, <div type='bibliography'>"
          
          bibliography_types:
            - type: "works_cited"
              description: "Works directly cited in text"
            - type: "further_reading"
              description: "Recommended but not cited works"
            - type: "sources"
              description: "Primary source list"
            - type: "discography"
              description: "Music recording references"
            - type: "filmography"
              description: "Film/video references"
            - type: "webography"
              description: "Web resource references"
          
          bibliography_entry_components:
            - component: "BIB.AUT"
              name: "Author"
              entity_type: "AGT"
              ontology: "dcterms:creator"
            - component: "BIB.TTL"
              name: "Title"
              entity_type: "WRK"
              ontology: "dcterms:title"
            - component: "BIB.DAT"
              name: "Date"
              entity_type: "TMP"
              ontology: "dcterms:date"
            - component: "BIB.PUB"
              name: "Publisher"
              entity_type: "GRP"
              ontology: "dcterms:publisher"
            - component: "BIB.PLC"
              name: "Place"
              entity_type: "TOP"
              ontology: "dcterms:spatial"
          
          entity_extraction_notes: |
            Bibliographies are PRE-STRUCTURED entity sources:
            
            1. EACH ENTRY contains:
               - Author(s) → AGENT entities
               - Title → WORK entity
               - Date → TEMPORAL entity
               - Publisher → GROUP entity
               - Place → TOPONYM entity
            
            2. CITATION PARSING:
               - Citation format indicates field boundaries
               - High confidence due to intentional structure            
        
        DOC.APP:
          name: "APPENDIX"
          code: "DOC.APP"
          definition: |
            Supplementary material at document end containing supporting
            data, extended discussions, or reference material too detailed
            for main text.
            
            Domain-agnostic applications:
            - Heritage: Document transcriptions, provenance records
            - Publishing: Data tables, extended methods, source texts
            - Academic: Statistical data, interview transcripts
            - Legal: Exhibits, supporting documents
            - Archives: Finding aid supplements, accession lists            
          
          ontology_mappings:
            primary_class: "bibo:DocumentPart"
            alternative_classes:
              - "schema:CreativeWork"
            linkml_mapping:
              class_uri: "bibo:DocumentPart"
          
          format_mappings:
            html:
              elements: [".appendix", "section[role='doc-appendix']"]
            tei:
              element: "<div type='appendix'>"
          
          appendix_types:
            - type: "data_appendix"
              description: "Statistical or tabular data"
            - type: "document_appendix"
              description: "Reproduced primary documents"
            - type: "technical_appendix"
              description: "Methods, algorithms, specifications"
            - type: "glossary_appendix"
              description: "Extended terminology (see also DOC.GLO)"
          
          appendix_properties:
            - property: "appendixLabel"
              uri: "schema:name"
              description: "Appendix letter/number (A, B, C or 1, 2, 3)"
            - property: "appendixTitle"
              uri: "dcterms:title"
              description: "Descriptive title"
        
        DOC.GLO:
          name: "GLOSSARY"
          code: "DOC.GLO"
          definition: |
            Alphabetical list of terms with definitions. Each entry is a
            CONCEPT with term (label) and definition (description).
            
            Domain-agnostic applications:
            - Heritage: Manuscript terminology, archaic word lists
            - Publishing: Technical glossaries, foreign word lists
            - Academic: Disciplinary terminology
            - Legal: Legal terms, statutory definitions
            - Archives: Archival terminology, provenance terms
            
            Glossaries are CONTROLLED VOCABULARY sources.            
          
          ontology_mappings:
            primary_class: "skos:ConceptScheme"
            alternative_classes:
              - "schema:DefinedTermSet"
              - "bibo:DocumentPart"
            linkml_mapping:
              class_uri: "skos:ConceptScheme"
          
          format_mappings:
            html:
              elements: [".glossary", "dl.glossary", "section[role='doc-glossary']"]
            tei:
              element: "<list type='gloss'>, <div type='glossary'>"
          
          glossary_entry_components:
            - component: "GLO.TRM"
              name: "Term"
              ontology: "skos:prefLabel"
            - component: "GLO.DEF"
              name: "Definition"
              ontology: "skos:definition"
            - component: "GLO.SYN"
              name: "Synonym"
              ontology: "skos:altLabel"
            - component: "GLO.REL"
              name: "Related Term"
              ontology: "skos:related"
        
        # ---------------------------------------------------------------------
        # COMMERCIAL AND BRANDING REGIONS
        # ---------------------------------------------------------------------
        
        DOC.ADV:
          name: "ADVERTISEMENT"
          code: "DOC.ADV"
          definition: |
            Commercial or promotional content within document. Important for
            historical documents where ads provide dating, pricing, business,
            and social context. Distinct from main editorial content.
            
            Domain-agnostic applications:
            - Heritage: Historical newspaper ads, trade catalog entries, broadside ads
            - Publishing: Book advertisements, periodical ads, classified sections
            - Archives: Commercial records, promotional materials
            - Web: Banner ads, sponsored content, promotional sections
            - Ephemera: Trade cards, handbills, promotional flyers
            
            Advertisements are RICH entity sources for historical research:
            business names, addresses, prices, products, and social attitudes.            
          
          ontology_mappings:
            primary_class: "schema:Advertisement"
            alternative_classes:
              - "crm:E73_Information_Object"
              - "bibo:DocumentPart"
            linkml_mapping:
              class_uri: "schema:Advertisement"
          
          format_mappings:
            html:
              elements: [".ad", ".advertisement", "[role='complementary'][aria-label*='sponsor']", "aside.ad"]
            pagexml:
              type: "TextRegion[@type='advertisement']"
            newspaper:
              note: "Common in historical newspaper digitization"
          
          advertisement_types:
            - type: "display_ad"
              description: "Large format advertisement with graphics"
            - type: "classified_ad"
              description: "Text-only small advertisement"
            - type: "trade_listing"
              description: "Business directory entry"
            - type: "prospectus"
              description: "Book/publication advertisement"
            - type: "patent_medicine"
              description: "Historical medical product ads"
              note: "Common in 19th century periodicals"
            - type: "auction_notice"
              description: "Sale or auction announcement"
            - type: "legal_notice"
              description: "Required public announcements"
          
          advertisement_properties:
            - property: "advertiser"
              uri: "schema:sponsor"
              description: "Business or person placing ad"
              entity_type: "GRP or AGT"
            - property: "product"
              uri: "schema:itemAdvertised"
              description: "Product or service advertised"
            - property: "businessAddress"
              uri: "schema:address"
              description: "Advertiser's address"
              entity_type: "TOP"
            - property: "price"
              uri: "schema:price"
              description: "Advertised price"
              entity_type: "QTY"
          
          entity_extraction_notes: |
            Historical advertisements are TREASURE TROVES:
            
            1. BUSINESS ENTITIES:
               - Business names → GROUP entities
               - Proprietor names → AGENT entities
               - Business addresses → TOPONYM entities
            
            2. HISTORICAL VALUE:
               - Dating evidence (product availability)
               - Pricing history
               - Business locations over time
               - Social/cultural attitudes
            
            3. PROVENANCE:
               - Distinguish ad claims from editorial claims
               - Ads have different authority level            
        
        DOC.LOG:
          name: "LOGO"
          code: "DOC.LOG"
          definition: |
            Visual identity marks: logos, mastheads, colophon marks, printer's
            devices, watermarks, seals, and brand identifiers. Important for
            attribution and provenance.
            
            Domain-agnostic applications:
            - Heritage: Printer's marks, publisher devices, watermarks, seals
            - Publishing: Publisher logos, journal mastheads, imprint marks
            - Archives: Institutional seals, letterhead logos
            - Web: Site logos, brand marks, favicons
            - Legal: Notary seals, official stamps, certification marks
            
            Logos identify PRODUCING AGENTS and provide provenance evidence.            
          
          ontology_mappings:
            primary_class: "schema:ImageObject"
            alternative_classes:
              - "crm:E37_Mark"
              - "crm:E73_Information_Object"
            linkml_mapping:
              class_uri: "schema:ImageObject"
              close_mappings:
                - "crm:E37_Mark"
          
          format_mappings:
            html:
              elements: [".logo", "header img.logo", "[role='banner'] img", ".masthead img"]
            pagexml:
              type: "GraphicRegion[@type='logo'], GraphicRegion[@type='decoration']"
          
          logo_types:
            - type: "publisher_logo"
              description: "Publisher's identifying mark"
            - type: "printer_device"
              description: "Historical printer's identifying mark"
              note: "Critical for incunabula identification"
            - type: "masthead"
              description: "Newspaper/periodical title banner"
            - type: "watermark"
              description: "Paper manufacturer's mark"
              note: "Used for paper dating and provenance"
            - type: "seal"
              description: "Official or personal seal impression"
              entity_type: "AGT or GRP"
            - type: "coat_of_arms"
              description: "Heraldic device"
              entity_type: "AGT or GRP"
            - type: "colophon_mark"
              description: "Decorative mark in colophon"
            - type: "ex_libris"
              description: "Bookplate or ownership mark"
              note: "Provenance evidence for ownership history"
          
          logo_properties:
            - property: "logoOwner"
              uri: "schema:creator"
              description: "Entity identified by logo"
              entity_type: "GRP or AGT"
            - property: "logoDescription"
              uri: "schema:description"
              description: "Visual description of mark"
            - property: "logoReference"
              uri: "schema:isBasedOn"
              description: "Reference to mark catalog/database"
          
          entity_extraction_notes: |
            Logos provide PROVENANCE evidence:
            
            1. ATTRIBUTION:
               - Printer's devices identify producer
               - Publisher logos identify publisher
               - Watermarks date paper production
            
            2. OWNERSHIP HISTORY:
               - Ex libris marks trace ownership
               - Seals indicate institutional provenance
            
            3. VISUAL ANALYSIS required:
               - May need image matching to logo databases
               - Heraldic interpretation for coats of arms            
        
        DOC.PGN:
          name: "PAGINATION"
          definition: |
            Page numbers, folio numbers, signature marks in printed/manuscript works.            
          
          ontology_mappings:
            primary_class: "crm:E42_Identifier"
          
          pagination_types:
            - type: "page_number"
              pagexml: "TextRegion[@type='page-number']"
              description: "Arabic or Roman numeral page number"
            - type: "folio"
              description: "Leaf number with recto/verso (e.g., 23r, 23v)"
            - type: "signature_mark"
              pagexml: "TextRegion[@type='signature-mark']"
              description: "Gathering/quire identifier in manuscripts"
            - type: "catch_word"
              pagexml: "TextRegion[@type='catch-word']"
              description: "Word at page bottom matching next page start"
        
        DOC.BLK:
          name: "BLOCK_QUOTE"
          definition: |
            Extended quotation from another source, typically indented or styled
            distinctly from surrounding text.            
          
          ontology_mappings:
            primary_class: "schema:Quotation"
            alternative_classes:
              - "crm:E33_Linguistic_Object"
          
          quote_properties:
            - property: "quotedFrom"
              uri: "glam:quotedFrom"
              range: "xsd:anyURI"
              owl_mapping: "prov:wasDerivedFrom"
              description: "Source of the quotation"
        
        # ---------------------------------------------------------------------
        # METADATA AND ADMINISTRATIVE REGIONS
        # ---------------------------------------------------------------------
        
        DOC.MTD:
          name: "METADATA_BLOCK"
          definition: |
            Region containing document metadata (author, date, keywords, etc.).
            High-value for entity extraction as claims are typically structured.
            
            Includes: HTML <head>, document properties, front matter, colophon.            
          
          ontology_mappings:
            primary_class: "dcterms:BibliographicResource"
            alternative_classes:
              - "schema:CreativeWork"
          
          metadata_block_types:
            - type: "front_matter"
              description: "Title page, copyright, dedication"
            - type: "back_matter"
              description: "Appendices, bibliography, colophon"
            - type: "colophon"
              description: "Production details (printer, date, place)"
              pagexml: "TextRegion[@type='colophon']"
            - type: "document_head"
              html: "head"
              description: "HTML metadata section"
        
        DOC.ANN:
          name: "ANNOTATION_REGION"
          definition: |
            Region containing annotations or markup added to document.
            Distinguished from original content for provenance tracking.
            
            Includes: Editorial additions, transcription notes, TEI annotations.            
          
          ontology_mappings:
            primary_class: "oa:Annotation"
            alternative_classes:
              - "crm:E13_Attribute_Assignment"
          
          annotation_properties:
            - property: "annotationBody"
              uri: "oa:hasBody"
              description: "Content of the annotation"
            - property: "annotationTarget"
              uri: "oa:hasTarget"
              description: "Region being annotated"
            - property: "annotator"
              uri: "oa:annotatedBy"
              owl_mapping: "prov:wasAttributedTo"
              description: "Agent who created annotation"
    
    # -------------------------------------------------------------------------
    # SEMANTIC ROLE ENUMERATION
    # -------------------------------------------------------------------------
    
    layout_semantic_roles:
      description: |
        Enumeration of semantic roles that document regions can play.
        A single region may have multiple roles.        
      
      roles:
        - role: "PRIMARY_CONTENT"
          code: "PRIM"
          description: "Main content bearing primary information"
          typical_regions: ["DOC.PAR", "DOC.HDR", "DOC.LST"]
        
        - role: "SUPPLEMENTARY"
          code: "SUPP"
          description: "Additional context or metadata"
          typical_regions: ["DOC.SDB", "DOC.FTN", "DOC.CAP", "DOC.APP"]
        
        - role: "NAVIGATIONAL"
          code: "NAV"
          description: "Aids document navigation"
          typical_regions: ["DOC.NAV", "DOC.PGN", "DOC.TOC", "DOC.IDX"]
        
        - role: "STRUCTURAL"
          code: "STRC"
          description: "Defines document structure"
          typical_regions: ["DOC.HDR", "DOC.TTP"]
        
        - role: "REFERENTIAL"
          code: "REF"
          description: "Points to other resources"
          typical_regions: ["DOC.FTN", "DOC.BLK", "DOC.BIB"]
        
        - role: "VISUAL"
          code: "VIS"
          description: "Non-textual visual content (images, maps, diagrams)"
          typical_regions: ["DOC.FIG", "DOC.GAL", "DOC.MAP", "DOC.LOG"]
        
        - role: "AUDIOVISUAL"
          code: "AV"
          description: "Time-based media content (audio, video)"
          typical_regions: ["DOC.AUD", "DOC.VID"]
        
        - role: "INTERACTIVE"
          code: "INT"
          description: "User-manipulable embedded content"
          typical_regions: ["DOC.EMB", "DOC.MAP"]
          note: "Interactive maps have both VIS and INT roles"
        
        - role: "METADATA"
          code: "META"
          description: "Document-level metadata"
          typical_regions: ["DOC.MTD", "DOC.COL", "DOC.TTP"]
        
        - role: "SPATIAL"
          code: "SPAT"
          description: "Geographic or spatial representation"
          typical_regions: ["DOC.MAP"]
          note: "Distinct from VISUAL; emphasizes coordinate/location semantics"
        
        - role: "FRONT_MATTER"
          code: "FRNT"
          description: "Preliminary material before main content"
          typical_regions: ["DOC.TTP", "DOC.DED", "DOC.TOC"]
        
        - role: "BACK_MATTER"
          code: "BACK"
          description: "Material following main content"
          typical_regions: ["DOC.BIB", "DOC.IDX", "DOC.APP", "DOC.GLO", "DOC.COL"]
        
        - role: "PARATEXTUAL"
          code: "PARA"
          description: "Content about the document itself (not subject matter)"
          typical_regions: ["DOC.TTP", "DOC.COL", "DOC.DED", "DOC.ADV"]
          note: "Genette's paratext concept - frames the main text"
        
        - role: "COMMERCIAL"
          code: "COMM"
          description: "Commercial or promotional content"
          typical_regions: ["DOC.ADV", "DOC.LOG"]
          note: "Distinguish from editorial content for authority assessment"
        
        - role: "LEXICAL"
          code: "LEX"
          description: "Vocabulary, terminology, and definitions"
          typical_regions: ["DOC.GLO", "DOC.IDX"]
  
  # ---------------------------------------------------------------------------
  # NESTED PROVENANCE MODEL
  # ---------------------------------------------------------------------------
  # Entity claims reference layout claims in a nested provenance structure.
  # ---------------------------------------------------------------------------
  
  nested_provenance_model:
    
    description: |
      Entity annotations within documents require TWO LAYERS of provenance:
      
      LAYER 1: LAYOUT CLAIM
      - Asserts that a document region has a particular semantic role
      - Has its own provenance chain (extraction method, confidence, timestamp)
      - May be uncertain (e.g., "80% confident this is marginalia, not sidebar")
      
      LAYER 2: ENTITY CLAIM (nested)
      - Asserts that an entity appears within the layout region
      - References Layer 1 layout claim as context
      - Has its own provenance chain for the entity extraction
      
      This separation enables:
      - Different models for layout vs. entity extraction
      - Independent validation of each layer
      - Reasoning about how layout uncertainty affects entity confidence
      - Aggregation of claims by layout context      
    
    layout_claim_schema:
      description: "Schema for layout/structural claims about document regions"
      
      required_fields:
        - field: "layout_claim_id"
          type: "URI"
          description: "Unique identifier for this layout claim"
          example: "https://example.org/claims/layout/doc1-region-5"
        
        - field: "region_type"
          type: "LayoutRegionCode"
          description: "Code from layout semantic ontology (DOC.HDR, DOC.SDB.MRG, etc.)"
          example: "DOC.SDB.MRG"
        
        - field: "region_path"
          type: "string"
          description: "XPath, JSONPath, or path to region"
          example: "//page:TextRegion[@id='r5']"
        
        - field: "source_document"
          type: "URI"
          description: "Document containing this region"
          example: "https://example.org/documents/manuscript_123.xml"
        
        - field: "document_format"
          type: "enum"
          values: ["PAGE-XML", "HTML", "JSON", "TEI-XML", "MARKDOWN", "PDF", "EPUB"]
        
        - field: "extraction_timestamp"
          type: "xsd:dateTime"
          owl_mapping: "prov:generatedAtTime"
        
        - field: "extraction_agent"
          type: "URI"
          description: "Model or agent that identified this region"
          owl_mapping: "prov:wasAttributedTo"
        
        - field: "confidence"
          type: "xsd:float"
          range: "[0.0, 1.0]"
          description: "Confidence in layout classification"
      
      optional_fields:
        - field: "parent_region"
          type: "URI"
          description: "Reference to containing region's layout claim"
        
        - field: "semantic_roles"
          type: "list[SemanticRoleCode]"
          description: "Roles this region plays (PRIM, SUPP, NAV, etc.)"
        
        - field: "governing_header"
          type: "URI"
          description: "Reference to header region that governs this content"
        
        - field: "layout_model_version"
          type: "string"
          description: "Version of layout extraction model used"
    
    entity_claim_with_layout_context:
      description: "Schema for entity claims that reference layout context"
      
      required_fields:
        - field: "entity_claim_id"
          type: "URI"
          description: "Unique identifier for this entity claim"
        
        - field: "entity_type"
          type: "EntityTypeCode"
          description: "Code from entity type ontology (AGT.PER, TOP.CTY, etc.)"
        
        - field: "entity_text"
          type: "string"
          description: "Surface form of entity mention"
          owl_mapping: "nif:anchorOf"
        
        - field: "span_start"
          type: "xsd:integer"
          owl_mapping: "nif:beginIndex"
        
        - field: "span_end"
          type: "xsd:integer"
          owl_mapping: "nif:endIndex"
        
        - field: "layout_context"
          type: "URI"
          description: "Reference to layout claim (Layer 1)"
          note: "CRITICAL: Links entity to its structural context"
        
        - field: "extraction_timestamp"
          type: "xsd:dateTime"
          owl_mapping: "prov:generatedAtTime"
        
        - field: "extraction_agent"
          type: "URI"
          owl_mapping: "prov:wasAttributedTo"
        
        - field: "confidence"
          type: "xsd:float"
          range: "[0.0, 1.0]"
      
      optional_fields:
        - field: "entity_link"
          type: "URI"
          description: "Link to knowledge base entity (Wikidata, etc.)"
          owl_mapping: "itsrdf:taIdentRef"
        
        - field: "entity_link_confidence"
          type: "xsd:float"
          description: "Confidence in entity linking (separate from NER confidence)"
        
        - field: "inherited_context"
          type: "object"
          description: "Context inherited from governing header"
          schema:
            temporal_scope: "xsd:string"
            spatial_scope: "xsd:string"
            topic_scope: "xsd:string"
    
    example_nested_claims:
      description: "Example showing nested layout and entity claims"
      
      layout_claim:
        layout_claim_id: "https://example.org/claims/layout/ms123-marginalia-1"
        region_type: "DOC.SDB.MRG"
        region_path: "//page:TextRegion[@id='r5'][@type='marginalia-left']"
        source_document: "https://example.org/documents/manuscript_123.xml"
        document_format: "PAGE-XML"
        extraction_timestamp: "2025-12-02T14:30:00Z"
        extraction_agent: "https://example.org/agents/layout-model-v2.1"
        confidence: 0.92
        semantic_roles: ["SUPP", "META"]
        parent_region: "https://example.org/claims/layout/ms123-page-1"
      
      entity_claim:
        entity_claim_id: "https://example.org/claims/entity/ms123-date-1"
        entity_type: "TMP.DAB"
        entity_text: "anno 1642"
        span_start: 12
        span_end: 21
        layout_context: "https://example.org/claims/layout/ms123-marginalia-1"
        extraction_timestamp: "2025-12-02T14:31:00Z"
        extraction_agent: "https://example.org/agents/ner-model-v3.0"
        confidence: 0.88
        inherited_context:
          note: "Date in marginalia - may indicate when reader annotated, not content date"
    
    example_wikipedia_infobox:
      description: |
        Complex real-world example: Extracting structured data from a Wikipedia
        infobox. Shows how layout uncertainty affects entity confidence and how
        multiple entities in one region share a common layout claim parent.        
      
      # First, the layout claim about the infobox region itself
      layout_claim:
        layout_claim_id: "https://example.org/claims/layout/wiki-rijksmuseum-infobox"
        region_type: "DOC.SDB.IBX"
        region_path: "//table[@class='infobox vcard']"
        source_document: "https://en.wikipedia.org/wiki/Rijksmuseum"
        document_format: "HTML"
        extraction_timestamp: "2025-12-02T16:00:00Z"
        extraction_agent: "https://example.org/agents/html-layout-detector-v1.0"
        confidence: 0.98
        semantic_roles: ["SUPP", "META"]
        parent_region: null
        note: |
          High confidence because Wikipedia infoboxes have consistent class names.
          This is SUPPLEMENTARY content (not main article) providing METADATA.          
      
      # Entity claims nested within this layout region
      entity_claims:
        - entity_claim_id: "https://example.org/claims/entity/rijksmuseum-founding-date"
          entity_type: "TMP.DAB"
          entity_text: "1800"
          span_start: 342
          span_end: 346
          layout_context: "https://example.org/claims/layout/wiki-rijksmuseum-infobox"
          extraction_timestamp: "2025-12-02T16:01:00Z"
          extraction_agent: "https://example.org/agents/ner-temporal-v2.0"
          confidence: 0.95
          entity_link: "https://www.wikidata.org/entity/Q190804"
          entity_link_confidence: 0.92
          property_path: "//table[@class='infobox vcard']//th[text()='Established']/following-sibling::td"
          property_label: "Established"
          note: |
            Property extracted from infobox row. The label "Established" provides
            semantic context that this is a founding date, not just any date.            
        
        - entity_claim_id: "https://example.org/claims/entity/rijksmuseum-location"
          entity_type: "TOP.CTY"
          entity_text: "Amsterdam"
          span_start: 128
          span_end: 137
          layout_context: "https://example.org/claims/layout/wiki-rijksmuseum-infobox"
          extraction_timestamp: "2025-12-02T16:01:00Z"
          extraction_agent: "https://example.org/agents/ner-spatial-v2.0"
          confidence: 0.99
          entity_link: "https://www.wikidata.org/entity/Q727"
          entity_link_confidence: 0.99
          property_path: "//table[@class='infobox vcard']//th[text()='Location']/following-sibling::td"
          property_label: "Location"
        
        - entity_claim_id: "https://example.org/claims/entity/rijksmuseum-name"
          entity_type: "GRP.ORG"
          entity_text: "Rijksmuseum"
          span_start: 0
          span_end: 11
          layout_context: "https://example.org/claims/layout/wiki-rijksmuseum-infobox"
          extraction_timestamp: "2025-12-02T16:01:00Z"
          extraction_agent: "https://example.org/agents/ner-org-v2.0"
          confidence: 0.99
          entity_link: "https://www.wikidata.org/entity/Q190804"
          entity_link_confidence: 0.99
          property_path: "//table[@class='infobox vcard']//caption"
          property_label: "infobox_title"
      
      provenance_chain_explanation: |
        This example demonstrates the TWO-LAYER provenance model:
        
        LAYER 1 (Layout Claim):
        - Claims the table is an infobox (DOC.SDB.IBX)
        - Has its own confidence (0.98) and agent (html-layout-detector-v1.0)
        - If layout is misclassified, ALL nested entity claims are affected
        
        LAYER 2 (Entity Claims):
        - Each entity references the layout claim via layout_context
        - Each entity has its OWN confidence separate from layout confidence
        - Entity linking confidence is SEPARATE from NER confidence
        
        Combined confidence calculation:
        - P(entity correct) = P(layout correct) × P(NER correct | layout)
        - P("1800" is Rijksmuseum founding) ≈ 0.98 × 0.95 = 0.931
        
        If layout claim is later corrected (e.g., this wasn't actually an infobox),
        all entity claims that referenced it can be invalidated or reprocessed.        
  
  # ---------------------------------------------------------------------------
  # Structural Context for Entity Clustering
  # ---------------------------------------------------------------------------
  structural_contexts:
    
    header_contexts:
      description: |
        Headers (h1-h6, TextRegion[@type='header'], JSON keys) establish scope
        for subsequent content. Entities in headers often:
        - Name the TOPIC of the section
        - Establish the SUBJECT of subsequent predicates
        - Define the CONTEXT for entity resolution        
      
      examples:
        - context: "## Board of Directors"
          implication: "Subsequent person names are likely board members"
        - context: "<heading>1580-1620: The Golden Age</heading>"
          implication: "Subsequent dates are within this period context"
        - context: "{ \"museums\": { \"amsterdam\": { ... } } }"
          implication: "Nested entities are Amsterdam museums"
      
      annotation_rules:
        - rule: "TAG_HEADER_ENTITIES"
          description: |
            Entities in headers MUST be tagged with structural role metadata:
            - header_level: 1-6 (depth in hierarchy)
            - governs_region: XPath/JSONPath to governed content
            - scope_type: TOPIC | TEMPORAL | SPATIAL | ORGANIZATIONAL            
        
        - rule: "PROPAGATE_HEADER_CONTEXT"
          description: |
            Claims about entities in governed content inherit header context:
            - A person in "Board of Directors" section gets implicit org:memberOf
            - A date in "1580-1620" section gets implicit temporal bounding            
    
    paragraph_contexts:
      description: |
        Paragraphs (p, TextRegion[@type='paragraph'], text values) contain
        the primary content. Entities in paragraphs:
        - Co-occur with other entities in same paragraph (potential relationships)
        - Fall under scope of governing headers
        - May reference entities from previous paragraphs (anaphora)        
      
      annotation_rules:
        - rule: "PARAGRAPH_CLUSTERING"
          description: |
            Entities within the same paragraph form a CO-OCCURRENCE CLUSTER.
            Co-occurrence is a weak signal of relationship, strengthened by:
            - Syntactic proximity (same sentence > adjacent sentences)
            - Shared predicate (both objects of same verb)
            - List membership (items in same enumeration)            
        
        - rule: "CROSS_PARAGRAPH_REFERENCE"
          description: |
            Pronouns and definite references ("the museum", "he") resolve to
            antecedents in prior paragraphs within the same section scope.
            Track paragraph index for resolution distance weighting.            
    
    list_contexts:
      description: |
        Lists (ul/ol, enumerations, bullet points) indicate PARALLEL STRUCTURE.
        List items share a common type or relationship to a parent concept.        
      
      annotation_rules:
        - rule: "LIST_ITEM_PARALLELISM"
          description: |
            Entities in sibling list items likely share:
            - Same entity TYPE (all museums, all dates, all people)
            - Same RELATIONSHIP to list parent (all "member of", all "located in")
            - Parallel STRUCTURE (if first item has date, others likely do too)            
        
        - rule: "LIST_HEADER_INHERITANCE"
          description: |
            List items inherit context from the text introducing the list.
            "The following museums are members:" + <li>Rijksmuseum</li>
            implies Rijksmuseum has membership relationship.            
    
    sidebar_contexts:
      description: |
        Sidebars, asides, marginalia, infoboxes contain SUPPLEMENTARY information.
        Entities in sidebars:
        - Provide metadata ABOUT the main content
        - May contain structured data (birth dates, locations, identifiers)
        - Have different reliability/authority than main narrative        
      
      annotation_rules:
        - rule: "SIDEBAR_METADATA_EXTRACTION"
          description: |
            Sidebar content often contains STRUCTURED CLAIMS suitable for
            direct property extraction:
            - Infobox fields → claim properties
            - Marginalia dates → temporal metadata
            - Caption text → aboutness relationships            
        
        - rule: "SIDEBAR_MAIN_LINKING"
          description: |
            Link sidebar entities to main content entities when co-referential.
            The sidebar "Born: 1606" links to the main text "Rembrandt".            
    
    caption_contexts:
      description: |
        Captions (figcaption, TextRegion[@type='caption']) describe VISUAL content.
        Entities in captions:
        - Describe depicted subjects (people, places, objects)
        - Provide dates/locations for the depicted scene
        - May differ from main text (image of one thing, text about another)        
      
      annotation_rules:
        - rule: "CAPTION_VISUAL_BINDING"
          description: |
            Caption entities are ABOUT the associated figure/image.
            This is distinct from main text co-occurrence.
            Use relationship: schema:about with target: figure URI.            
        
        - rule: "CAPTION_PROVENANCE"
          description: |
            Captions may have different authorship/dates than main text.
            Track caption-specific provenance when available.            
  
  # ---------------------------------------------------------------------------
  # Format-Specific Path Conventions
  # ---------------------------------------------------------------------------
  format_path_conventions:
    
    page_xml:
      description: |
        PAGE-XML (used for historical manuscript transcription) organizes content
        into TextRegions with type attributes. Critical for manuscript NER.        
      
      namespace: "http://schema.primaresearch.org/PAGE/gts/pagecontent/2019-07-15"
      prefix: "page"
      
      text_region_types:
        - type: "header"
          semantic_role: "section_heading"
          path_example: "//page:TextRegion[@type='header'][1]"
        - type: "paragraph"
          semantic_role: "body_content"
          path_example: "//page:TextRegion[@type='paragraph'][3]"
        - type: "marginalia-left"
          semantic_role: "supplementary_note"
          path_example: "//page:TextRegion[@type='marginalia-left'][1]"
        - type: "marginalia-right"
          semantic_role: "supplementary_note"
          path_example: "//page:TextRegion[@type='marginalia-right'][1]"
        - type: "caption"
          semantic_role: "figure_description"
          path_example: "//page:TextRegion[@type='caption'][1]"
        - type: "page-number"
          semantic_role: "pagination"
          path_example: "//page:TextRegion[@type='page-number'][1]"
        - type: "signature-mark"
          semantic_role: "gathering_identifier"
          path_example: "//page:TextRegion[@type='signature-mark'][1]"
        - type: "catch-word"
          semantic_role: "gathering_continuity"
          path_example: "//page:TextRegion[@type='catch-word'][1]"
        - type: "table"
          semantic_role: "structured_data"
          path_example: "//page:TextRegion[@type='table'][1]"
      
      annotation_output:
        description: "Include PAGE-XML path in entity provenance"
        example:
          entity: "Rembrandt"
          span_start: 145
          span_end: 154
          page_xml_path: "//page:Page[@imageFilename='folio_23r.jpg']/page:TextRegion[@id='r1']/page:TextLine[@id='l1']/page:Word[@id='w3']"
          text_region_type: "paragraph"
          text_region_id: "r1"
    
    html:
      description: |
        HTML documents use semantic elements for structure. Critical for web NER.        
      
      semantic_elements:
        headers:
          elements: ["h1", "h2", "h3", "h4", "h5", "h6"]
          semantic_role: "section_heading"
          hierarchy: "h1 > h2 > h3 > h4 > h5 > h6"
        
        body_content:
          elements: ["p", "div", "span", "article", "section"]
          semantic_role: "primary_content"
        
        supplementary:
          elements: ["aside", "nav", "footer", "header"]
          semantic_role: "supplementary_content"
        
        figures:
          elements: ["figure", "figcaption", "img"]
          semantic_role: "visual_content"
        
        lists:
          elements: ["ul", "ol", "li", "dl", "dt", "dd"]
          semantic_role: "enumerated_content"
        
        tables:
          elements: ["table", "thead", "tbody", "tr", "th", "td"]
          semantic_role: "structured_data"
        
        metadata:
          elements: ["meta", "title", "head"]
          semantic_role: "document_metadata"
      
      xpath_conventions:
        - pattern: "//article/section[2]/h2[1]"
          description: "Second section's heading in article"
        - pattern: "//table[@class='infobox']//td[contains(text(),'Born')]"
          description: "Birth date cell in Wikipedia-style infobox"
        - pattern: "//aside//a[@href]"
          description: "Links in sidebar content"
      
      annotation_output:
        example:
          entity: "Rijksmuseum"
          span_start: 2341
          span_end: 2352
          xpath: "/html/body/main/article/section[3]/p[2]"
          css_selector: "article > section:nth-child(3) > p:nth-child(2)"
          semantic_context: "body_content"
          parent_header: "Dutch Museums"
          parent_header_path: "/html/body/main/article/section[3]/h2"
    
    json:
      description: |
        JSON documents use key paths for structure. Critical for API response NER.        
      
      path_notation: "JSONPath (RFC 9535) or dot notation"
      
      common_patterns:
        - pattern: "$.results[*].name"
          description: "Name field in array of results"
        - pattern: "$.data.institution.address.city"
          description: "Nested city within institution data"
        - pattern: "$['@context']"
          description: "JSON-LD context (for namespace resolution)"
        - pattern: "$.metadata.creator"
          description: "Document creator in metadata block"
      
      semantic_inference:
        description: |
          JSON keys often encode semantic roles:
          - "name", "title", "label" → entity names
          - "date", "created", "founded" → temporal properties
          - "location", "address", "place" → spatial properties
          - "author", "creator", "owner" → agent relationships
          - "type", "category", "class" → classification          
      
      annotation_output:
        example:
          entity: "Amsterdam"
          json_path: "$.data.museums[0].location.city"
          array_index: 0
          parent_key: "location"
          root_key: "museums"
    
    plain_text:
      description: |
        Plain text lacks structural markup. Use line/paragraph detection heuristics.        
      
      structure_detection:
        - heuristic: "BLANK_LINE_PARAGRAPH"
          description: "Two consecutive newlines indicate paragraph break"
        - heuristic: "INDENTATION_STRUCTURE"
          description: "Consistent indentation may indicate hierarchy"
        - heuristic: "CAPITALIZATION_HEADERS"
          description: "ALL CAPS or Title Case lines may be headers"
        - heuristic: "ENUMERATION_LISTS"
          description: "Lines starting with numbers/bullets are list items"
      
      annotation_output:
        description: "Use character offsets with paragraph/line indices"
        example:
          entity: "Dr. Jan de Wit"
          char_start: 1456
          char_end: 1470
          line_number: 23
          paragraph_index: 5
          inferred_section: "unknown"
  
  # ---------------------------------------------------------------------------
  # Entity Clustering by Structural Path
  # ---------------------------------------------------------------------------
  clustering_strategies:
    
    path_prefix_clustering:
      description: |
        Entities sharing a common path prefix belong to the same structural unit
        and should be clustered for co-reference resolution and relationship extraction.        
      
      algorithm:
        - step: 1
          action: "Extract path for each entity annotation"
        - step: 2
          action: "Compute longest common prefix (LCP) for entity pairs"
        - step: 3
          action: "Cluster entities where LCP depth >= threshold"
        - step: 4
          action: "Within clusters, resolve co-references before cross-cluster"
      
      example:
        entities:
          - name: "Rembrandt"
            path: "/article/section[2]/p[1]"
          - name: "Saskia"
            path: "/article/section[2]/p[1]"
          - name: "the painter"
            path: "/article/section[2]/p[2]"
          - name: "Vermeer"
            path: "/article/section[3]/p[1]"
        
        clusters:
          - cluster_id: 1
            path_prefix: "/article/section[2]"
            entities: ["Rembrandt", "Saskia", "the painter"]
            note: "Co-reference: 'the painter' likely refers to Rembrandt"
          - cluster_id: 2
            path_prefix: "/article/section[3]"
            entities: ["Vermeer"]
            note: "Separate section, no co-reference with cluster 1"
    
    hierarchical_inheritance:
      description: |
        Properties from ancestor nodes propagate to descendant entities.
        Header entities establish context inherited by paragraph entities.        
      
      inheritance_rules:
        - rule: "TEMPORAL_SCOPE"
          description: "Date range in header bounds dates in body"
          example:
            header: "The 17th Century (1600-1699)"
            body_date: "1642"
            inherited_context: "temporal_scope: 1600-1699"
        
        - rule: "SPATIAL_SCOPE"
          description: "Location in header provides default for body entities"
          example:
            header: "Museums in Amsterdam"
            body_org: "Rijksmuseum"
            inherited_context: "default_location: Amsterdam"
        
        - rule: "TOPIC_SCOPE"
          description: "Subject in header provides aboutness for body"
          example:
            header: "Rembrandt van Rijn"
            body_text: "He painted The Night Watch"
            inherited_context: "pronoun_antecedent: Rembrandt van Rijn"
    
    cross_reference_resolution:
      description: |
        Resolve pronouns and definite references using structural proximity.        
      
      resolution_priority:
        - priority: 1
          scope: "same_sentence"
          description: "Check for antecedent in same sentence"
        - priority: 2
          scope: "same_paragraph"
          description: "Check preceding sentences in paragraph"
        - priority: 3
          scope: "same_section"
          description: "Check preceding paragraphs in section"
        - priority: 4
          scope: "section_header"
          description: "Check governing section header"
        - priority: 5
          scope: "document"
          description: "Check document-level prominent entities"
      
      example:
        text: |
          ## Rembrandt van Rijn
          
          The Dutch painter was born in Leiden. He moved to Amsterdam in 1631.
          His most famous work is The Night Watch.          
        
        resolutions:
          - pronoun: "The Dutch painter"
            antecedent: "Rembrandt van Rijn"
            resolution_scope: "section_header"
          - pronoun: "He"
            antecedent: "'The Dutch painter' -> 'Rembrandt van Rijn'"
            resolution_scope: "same_paragraph"
          - pronoun: "His"
            antecedent: "'He' -> 'Rembrandt van Rijn'"
            resolution_scope: "same_paragraph"
  
  # ---------------------------------------------------------------------------
  # Provenance Path Requirements
  # ---------------------------------------------------------------------------
  provenance_requirements:
    
    mandatory_path_fields:
      description: |
        ALL entity annotations MUST include path information for provenance.
        This enables verification, reproducibility, and precise citation.        
      
      required_fields:
        - field: "source_document_uri"
          type: "URI"
          description: "Identifier for the source document"
          example: "https://example.org/documents/manuscript_123.xml"
        
        - field: "document_format"
          type: "enum"
          values: ["PAGE-XML", "HTML", "JSON", "TEI-XML", "PLAIN_TEXT", "PDF"]
          description: "Format of source document (determines path syntax)"
        
        - field: "structural_path"
          type: "string"
          description: "XPath, JSONPath, or character offset path to entity"
          examples:
            - "//page:TextRegion[@id='r5']/page:TextLine[@id='l3']"
            - "/html/body/article/section[2]/p[1]"
            - "$.data.institutions[3].name"
            - "char:1456-1470;line:23;para:5"
        
        - field: "structural_context"
          type: "enum"
          values: 
            - "header"
            - "paragraph"
            - "list_item"
            - "table_cell"
            - "caption"
            - "sidebar"
            - "marginalia"
            - "footnote"
            - "metadata"
          description: "Semantic role of containing structure"
        
        - field: "governing_header"
          type: "object"
          optional: true
          description: "Path and text of governing header (if applicable)"
          schema:
            path: "string"
            text: "string"
            level: "integer"
    
    nif_context_alignment:
      description: |
        Path information aligns with NIF Context for annotation interchange.        
      
      nif_properties:
        - property: "nif:beginIndex"
          maps_to: "span_start (character offset)"
        - property: "nif:endIndex"
          maps_to: "span_end (character offset)"
        - property: "nif:sourceUrl"
          maps_to: "source_document_uri"
        - property: "nif:wasConvertedFrom"
          maps_to: "structural_path (as provenance)"
      
      example_nif_output:
        "@context": "http://persistence.uni-leipzig.org/nlp2rdf/contexts/nif-2.0.json"
        "@id": "https://example.org/doc#char=145,154"
        "@type": "nif:String"
        "nif:anchorOf": "Rembrandt"
        "nif:beginIndex": 145
        "nif:endIndex": 154
        "nif:referenceContext": "https://example.org/doc#char=0,5000"
        "glam:structuralPath": "//page:TextRegion[@id='r1']/page:TextLine[@id='l1']"
        "glam:structuralContext": "paragraph"
        "glam:governingHeader":
          path: "//page:TextRegion[@id='header1']"
          text: "Dutch Golden Age Artists"
          level: 2

# =============================================================================
# SECTION 16: RELATIONSHIP ANNOTATIONS
# =============================================================================
# 
# This section defines how entities relate to each other through typed predicates.
# Relationships are first-class annotations with their own provenance, enabling
# knowledge graph construction from annotated documents.
#
# Ontological foundations:
#   - CIDOC-CRM: CRMbase properties (P1-P198) for cultural heritage
#   - RDF/OWL: Standard predicate semantics
#   - Schema.org: Web-friendly relationship vocabulary
#   - FRBR/FRBRoo: Work-Expression-Manifestation-Item relationships
#   - RiC-O: Archival relationship properties
#   - Bio CRM: Biographical relationships (birth, death, marriage, etc.)
#   - Wikidata: P-properties for broad coverage
#
# Design principles:
#   1. Relationships are DIRECTIONAL (subject → predicate → object)
#   2. Relationships have PROVENANCE (who asserted, when, confidence)
#   3. Relationships can be TEMPORAL (valid during time period)
#   4. Relationships can be NESTED (relationship about relationship)
#   5. Relationships support REIFICATION (statements about statements)
# =============================================================================

relationship_annotations:
  
  description: |
    Relationship annotations capture semantic links between entities within and
    across documents. Unlike entity annotations which identify spans of text,
    relationship annotations connect two or more entity annotations via typed
    predicates drawn from established ontologies.
    
    Key distinction from coreference:
    - COREFERENCE: Two mentions refer to the SAME entity (identity)
    - RELATIONSHIP: Two DIFFERENT entities have a semantic connection
    
    Example:
    - Coreference: "Rembrandt" and "the painter" → same person
    - Relationship: "Rembrandt" created "The Night Watch" → two entities linked    
  
  namespaces:
    # Core relationship vocabularies
    crm: "http://www.cidoc-crm.org/cidoc-crm/"
    crmsci: "http://www.cidoc-crm.org/extensions/crmsci/"
    frbroo: "http://iflastandards.info/ns/fr/frbr/frbroo/"
    rico: "https://www.ica.org/standards/RiC/ontology#"
    rel: "http://purl.org/vocab/relationship/"
    bio: "http://purl.org/vocab/bio/0.1/"
    schema: "http://schema.org/"
    wdt: "http://www.wikidata.org/prop/direct/"
    dbo: "http://dbpedia.org/ontology/"
    gndo: "https://d-nb.info/standards/elementset/gnd#"
    edm: "http://www.europeana.eu/schemas/edm/"
    dc: "http://purl.org/dc/elements/1.1/"
    dcterms: "http://purl.org/dc/terms/"
    prov: "http://www.w3.org/ns/prov#"
    oa: "http://www.w3.org/ns/oa#"
    rdf: "http://www.w3.org/1999/02/22-rdf-syntax-ns#"
    rdfs: "http://www.w3.org/2000/01/rdf-schema#"
    owl: "http://www.w3.org/2002/07/owl#"
    skos: "http://www.w3.org/2004/02/skos/core#"
    time: "http://www.w3.org/2006/time#"
    geo: "http://www.opengis.net/ont/geosparql#"

  # ---------------------------------------------------------------------------
  # RELATIONSHIP HYPERNYM TAXONOMY
  # ---------------------------------------------------------------------------
  
  relationship_hypernyms:
    description: |
      Top-level categorization of relationship types. Each hypernym groups
      semantically related predicates and maps to established ontologies.      
    
    hypernyms:
      
      # -------------------------------------------------------------------------
      # REL.CRE - Creation/Production Relationships
      # -------------------------------------------------------------------------
      - code: "REL.CRE"
        label: "Creation"
        description: |
          Relationships involving the creation, production, or authorship of
          works, objects, or expressions. Central to cultural heritage.          
        ontology_mappings:
          - uri: "crm:P14_carried_out_by"
            label: "carried out by"
            inverse: "crm:P14i_performed"
          - uri: "crm:P94_has_created"
            label: "has created"
            inverse: "crm:P94i_was_created_by"
          - uri: "schema:creator"
            label: "creator"
          - uri: "schema:author"
            label: "author"
          - uri: "dc:creator"
            label: "creator"
          - uri: "frbroo:R17_created"
            label: "created (FRBR Work)"
        
        hyponyms:
          - code: "REL.CRE.AUT"
            label: "Authored"
            description: "Agent authored a textual work"
            predicates:
              - uri: "schema:author"
              - uri: "dcterms:creator"
              - uri: "wdt:P50"
                wikidata_label: "author"
            domain: ["AGT.PER", "AGT.GRP"]
            range: ["WRK.TXT", "WRK.WRK"]
            
          - code: "REL.CRE.ART"
            label: "Created (Artistic)"
            description: "Agent created an artistic work"
            predicates:
              - uri: "crm:P14_carried_out_by"
              - uri: "wdt:P170"
                wikidata_label: "creator"
            domain: ["AGT.PER", "AGT.GRP"]
            range: ["WRK.VIS", "WRK.WRK"]
            
          - code: "REL.CRE.COM"
            label: "Composed"
            description: "Agent composed a musical work"
            predicates:
              - uri: "schema:composer"
              - uri: "wdt:P86"
                wikidata_label: "composer"
            domain: ["AGT.PER"]
            range: ["WRK.MUS"]
            
          - code: "REL.CRE.DIR"
            label: "Directed"
            description: "Agent directed a performance or film"
            predicates:
              - uri: "schema:director"
              - uri: "wdt:P57"
                wikidata_label: "director"
            domain: ["AGT.PER"]
            range: ["WRK.PER", "WRK.CIN"]
            
          - code: "REL.CRE.EDT"
            label: "Edited"
            description: "Agent edited a work"
            predicates:
              - uri: "schema:editor"
              - uri: "wdt:P98"
                wikidata_label: "editor"
            domain: ["AGT.PER", "AGT.GRP"]
            range: ["WRK.TXT", "WRK.WRK"]
            
          - code: "REL.CRE.TRN"
            label: "Translated"
            description: "Agent translated a work"
            predicates:
              - uri: "schema:translator"
              - uri: "wdt:P655"
                wikidata_label: "translator"
            domain: ["AGT.PER"]
            range: ["WRK.TXT"]
            
          - code: "REL.CRE.ILL"
            label: "Illustrated"
            description: "Agent illustrated a work"
            predicates:
              - uri: "schema:illustrator"
              - uri: "wdt:P110"
                wikidata_label: "illustrator"
            domain: ["AGT.PER"]
            range: ["WRK.TXT", "WRK.VIS"]
            
          - code: "REL.CRE.PHO"
            label: "Photographed"
            description: "Agent photographed (captured image)"
            predicates:
              - uri: "schema:photographer"
              - uri: "wdt:P4876"
                wikidata_label: "photographer"
            domain: ["AGT.PER"]
            range: ["WRK.VIS"]
            
          - code: "REL.CRE.DES"
            label: "Designed"
            description: "Agent designed an object or building"
            predicates:
              - uri: "wdt:P287"
                wikidata_label: "designed by"
            domain: ["AGT.PER", "AGT.GRP"]
            range: ["WRK.OBJ", "TOP.BLT"]
            
          - code: "REL.CRE.INV"
            label: "Invented"
            description: "Agent invented something"
            predicates:
              - uri: "wdt:P61"
                wikidata_label: "discoverer or inventor"
            domain: ["AGT.PER", "AGT.GRP"]
            range: ["WRK.OBJ", "WRK.WRK"]

      # -------------------------------------------------------------------------
      # REL.TMP - Temporal Relationships
      # -------------------------------------------------------------------------
      - code: "REL.TMP"
        label: "Temporal"
        description: |
          Relationships with temporal semantics - events, periods, sequences.
          Integrates with CIDOC-CRM temporal entities and TimeML.          
        ontology_mappings:
          - uri: "crm:P4_has_time-span"
            label: "has time-span"
          - uri: "time:hasTime"
            label: "has time"
          - uri: "edm:happenedAt"
            label: "happened at (time)"
        
        hyponyms:
          - code: "REL.TMP.DUR"
            label: "During"
            description: "Entity existed/occurred during time period"
            predicates:
              - uri: "crm:P4_has_time-span"
              - uri: "time:inside"
            domain: ["*"]
            range: ["TMP.DUR", "TMP.RNG"]
            
          - code: "REL.TMP.BEF"
            label: "Before"
            description: "Entity/event occurred before another"
            predicates:
              - uri: "time:before"
              - uri: "crm:P120_occurs_before"
            domain: ["TMP.*", "EVT.*"]
            range: ["TMP.*", "EVT.*"]
            
          - code: "REL.TMP.AFT"
            label: "After"
            description: "Entity/event occurred after another"
            predicates:
              - uri: "time:after"
              - uri: "crm:P120i_occurs_after"
            domain: ["TMP.*", "EVT.*"]
            range: ["TMP.*", "EVT.*"]
            
          - code: "REL.TMP.CON"
            label: "Contemporaneous"
            description: "Entities/events overlapped in time"
            predicates:
              - uri: "time:intervalOverlaps"
              - uri: "edm:isRelatedTo"
            domain: ["*"]
            range: ["*"]
            
          - code: "REL.TMP.BGN"
            label: "Began"
            description: "Entity/event started at time"
            predicates:
              - uri: "time:hasBeginning"
              - uri: "crm:P82a_begin_of_the_begin"
            domain: ["*"]
            range: ["TMP.DAB", "TMP.RNG"]
            
          - code: "REL.TMP.END"
            label: "Ended"
            description: "Entity/event ended at time"
            predicates:
              - uri: "time:hasEnd"
              - uri: "crm:P82b_end_of_the_end"
            domain: ["*"]
            range: ["TMP.DAB", "TMP.RNG"]

      # -------------------------------------------------------------------------
      # REL.SPA - Spatial Relationships
      # -------------------------------------------------------------------------
      - code: "REL.SPA"
        label: "Spatial"
        description: |
          Relationships with spatial/geographic semantics. Uses GeoSPARQL
          topological relations and CIDOC-CRM place properties.          
        ontology_mappings:
          - uri: "crm:P53_has_former_or_current_location"
            label: "has former or current location"
          - uri: "geo:sfWithin"
            label: "within (spatial)"
          - uri: "schema:location"
            label: "location"
        
        hyponyms:
          - code: "REL.SPA.LOC"
            label: "Located In"
            description: "Entity is/was located in place"
            predicates:
              - uri: "crm:P53_has_former_or_current_location"
              - uri: "schema:location"
              - uri: "wdt:P276"
                wikidata_label: "location"
            domain: ["*"]
            range: ["TOP.*", "GEO.*"]
            
          - code: "REL.SPA.WTH"
            label: "Within"
            description: "Place is spatially within another place"
            predicates:
              - uri: "geo:sfWithin"
              - uri: "crm:P89_falls_within"
              - uri: "wdt:P131"
                wikidata_label: "located in administrative entity"
            domain: ["TOP.*", "GEO.*"]
            range: ["TOP.*", "GEO.*"]
            
          - code: "REL.SPA.CON"
            label: "Contains"
            description: "Place spatially contains another"
            predicates:
              - uri: "geo:sfContains"
              - uri: "crm:P89i_contains"
            domain: ["TOP.*", "GEO.*"]
            range: ["TOP.*", "GEO.*"]
            
          - code: "REL.SPA.ADJ"
            label: "Adjacent To"
            description: "Place is adjacent to another"
            predicates:
              - uri: "geo:sfTouches"
              - uri: "wdt:P47"
                wikidata_label: "shares border with"
            domain: ["TOP.*", "GEO.*"]
            range: ["TOP.*", "GEO.*"]
            
          - code: "REL.SPA.NER"
            label: "Near"
            description: "Entity is near a place (fuzzy proximity)"
            predicates:
              - uri: "schema:geo"
              - uri: "crm:P168_place_is_defined_by"
            domain: ["*"]
            range: ["TOP.*", "GEO.*"]
            
          - code: "REL.SPA.ORG"
            label: "Origin"
            description: "Entity originates from place"
            predicates:
              - uri: "schema:birthPlace"
              - uri: "wdt:P19"
                wikidata_label: "place of birth"
              - uri: "wdt:P495"
                wikidata_label: "country of origin"
            domain: ["AGT.PER", "WRK.*"]
            range: ["TOP.*"]

      # -------------------------------------------------------------------------
      # REL.SOC - Social Relationships
      # -------------------------------------------------------------------------
      - code: "REL.SOC"
        label: "Social"
        description: |
          Interpersonal and organizational relationships. Uses RELATIONSHIP
          vocabulary, Bio ontology, and social network predicates.          
        ontology_mappings:
          - uri: "rel:participantIn"
            label: "participant in"
          - uri: "bio:relationship"
            label: "relationship (bio)"
          - uri: "schema:knows"
            label: "knows"
        
        hyponyms:
          - code: "REL.SOC.FAM"
            label: "Family"
            description: "Family relationship between persons"
            predicates:
              - uri: "rel:parentOf"
              - uri: "rel:childOf"
              - uri: "rel:siblingOf"
              - uri: "rel:spouseOf"
              - uri: "wdt:P22"
                wikidata_label: "father"
              - uri: "wdt:P25"
                wikidata_label: "mother"
              - uri: "wdt:P26"
                wikidata_label: "spouse"
              - uri: "wdt:P40"
                wikidata_label: "child"
            domain: ["AGT.PER"]
            range: ["AGT.PER"]
            sub_relations:
              - code: "REL.SOC.FAM.PAR"
                label: "Parent Of"
                predicates: ["rel:parentOf", "wdt:P22", "wdt:P25"]
              - code: "REL.SOC.FAM.CHD"
                label: "Child Of"
                predicates: ["rel:childOf", "wdt:P40"]
              - code: "REL.SOC.FAM.SIB"
                label: "Sibling Of"
                predicates: ["rel:siblingOf", "wdt:P3373"]
              - code: "REL.SOC.FAM.SPO"
                label: "Spouse Of"
                predicates: ["rel:spouseOf", "wdt:P26"]
              - code: "REL.SOC.FAM.GRP"
                label: "Grandparent Of"
                predicates: ["wdt:P1038"]
            
          - code: "REL.SOC.PRO"
            label: "Professional"
            description: "Professional/occupational relationship"
            predicates:
              - uri: "rel:collaboratesWith"
              - uri: "schema:colleague"
              - uri: "wdt:P1066"
                wikidata_label: "student of"
              - uri: "wdt:P184"
                wikidata_label: "doctoral advisor"
            domain: ["AGT.PER"]
            range: ["AGT.PER", "GRP.*"]
            sub_relations:
              - code: "REL.SOC.PRO.STU"
                label: "Student Of"
                predicates: ["wdt:P1066"]
              - code: "REL.SOC.PRO.TCH"
                label: "Teacher Of"
                predicates: ["wdt:P802"]
              - code: "REL.SOC.PRO.COL"
                label: "Colleague Of"
                predicates: ["schema:colleague"]
              - code: "REL.SOC.PRO.MNT"
                label: "Mentored By"
                predicates: ["wdt:P184"]
            
          - code: "REL.SOC.MEM"
            label: "Membership"
            description: "Agent is member of group"
            predicates:
              - uri: "schema:memberOf"
              - uri: "crm:P107i_is_current_or_former_member_of"
              - uri: "wdt:P463"
                wikidata_label: "member of"
            domain: ["AGT.PER", "GRP.*"]
            range: ["GRP.*"]
            
          - code: "REL.SOC.EMP"
            label: "Employment"
            description: "Agent employed by organization"
            predicates:
              - uri: "schema:worksFor"
              - uri: "wdt:P108"
                wikidata_label: "employer"
            domain: ["AGT.PER"]
            range: ["GRP.ORG", "GRP.COR", "GRP.GOV"]
            
          - code: "REL.SOC.AFF"
            label: "Affiliation"
            description: "Agent affiliated with organization"
            predicates:
              - uri: "schema:affiliation"
              - uri: "wdt:P1416"
                wikidata_label: "affiliation"
            domain: ["AGT.PER"]
            range: ["GRP.*"]
            
          - code: "REL.SOC.KNO"
            label: "Knows"
            description: "Agent knows another agent"
            predicates:
              - uri: "schema:knows"
              - uri: "rel:knowsOf"
            domain: ["AGT.PER"]
            range: ["AGT.PER"]
            
          - code: "REL.SOC.PAT"
            label: "Patronage"
            description: "Patron-artist/client relationship"
            predicates:
              - uri: "wdt:P88"
                wikidata_label: "commissioned by"
            domain: ["AGT.PER", "GRP.*"]
            range: ["AGT.PER", "GRP.*"]

      # -------------------------------------------------------------------------
      # REL.ORG - Organizational Relationships
      # -------------------------------------------------------------------------
      - code: "REL.ORG"
        label: "Organizational"
        description: |
          Relationships between organizations, including hierarchy,
          succession, and structural connections.          
        ontology_mappings:
          - uri: "schema:parentOrganization"
            label: "parent organization"
          - uri: "rico:isSuccessorOf"
            label: "is successor of"
          - uri: "crm:P107_has_current_or_former_member"
            label: "has member"
        
        hyponyms:
          - code: "REL.ORG.PAR"
            label: "Parent Organization"
            description: "Organization is parent of another"
            predicates:
              - uri: "schema:parentOrganization"
              - uri: "wdt:P749"
                wikidata_label: "parent organization"
            domain: ["GRP.*"]
            range: ["GRP.*"]
            
          - code: "REL.ORG.SUB"
            label: "Subsidiary"
            description: "Organization is subsidiary of another"
            predicates:
              - uri: "schema:subOrganization"
              - uri: "wdt:P355"
                wikidata_label: "subsidiary"
            domain: ["GRP.*"]
            range: ["GRP.*"]
            
          - code: "REL.ORG.SUC"
            label: "Successor"
            description: "Organization succeeded another"
            predicates:
              - uri: "rico:isSuccessorOf"
              - uri: "wdt:P1366"
                wikidata_label: "replaced by"
            domain: ["GRP.*"]
            range: ["GRP.*"]
            
          - code: "REL.ORG.PRE"
            label: "Predecessor"
            description: "Organization preceded another"
            predicates:
              - uri: "rico:hasPredecessor"
              - uri: "wdt:P1365"
                wikidata_label: "replaces"
            domain: ["GRP.*"]
            range: ["GRP.*"]
            
          - code: "REL.ORG.MRG"
            label: "Merged Into"
            description: "Organization merged into another"
            predicates:
              - uri: "rico:isAssociatedWithEvent"
              - uri: "wdt:P7888"
                wikidata_label: "merged into"
            domain: ["GRP.*"]
            range: ["GRP.*"]
            
          - code: "REL.ORG.SPL"
            label: "Split From"
            description: "Organization split from another"
            predicates:
              - uri: "wdt:P807"
                wikidata_label: "separated from"
            domain: ["GRP.*"]
            range: ["GRP.*"]
            
          - code: "REL.ORG.FND"
            label: "Founded By"
            description: "Organization founded by agent"
            predicates:
              - uri: "schema:founder"
              - uri: "wdt:P112"
                wikidata_label: "founded by"
            domain: ["GRP.*"]
            range: ["AGT.PER", "GRP.*"]

      # -------------------------------------------------------------------------
      # REL.WRK - Work/Expression Relationships (FRBR)
      # -------------------------------------------------------------------------
      - code: "REL.WRK"
        label: "Work Relations"
        description: |
          Relationships between works, expressions, manifestations, and items
          following the FRBR model. Critical for bibliographic and archival data.          
        ontology_mappings:
          - uri: "frbroo:R3_is_realised_in"
            label: "is realised in (Work→Expression)"
          - uri: "frbroo:R4_carriers_provided_by"
            label: "carriers provided by (Expression→Manifestation)"
          - uri: "schema:isBasedOn"
            label: "is based on"
        
        hyponyms:
          - code: "REL.WRK.EXP"
            label: "Expression Of"
            description: "Expression realizes a Work"
            predicates:
              - uri: "frbroo:R3i_realises"
              - uri: "wdt:P629"
                wikidata_label: "edition or translation of"
            domain: ["WRK.EXP"]
            range: ["WRK.WRK"]
            
          - code: "REL.WRK.MAN"
            label: "Manifestation Of"
            description: "Manifestation embodies an Expression"
            predicates:
              - uri: "frbroo:R4i_comprises_carriers_of"
            domain: ["WRK.MAN"]
            range: ["WRK.EXP"]
            
          - code: "REL.WRK.ITM"
            label: "Item Of"
            description: "Item exemplifies a Manifestation"
            predicates:
              - uri: "frbroo:R7i_is_example_of"
            domain: ["WRK.ITM"]
            range: ["WRK.MAN"]
            
          - code: "REL.WRK.DRV"
            label: "Derived From"
            description: "Work derived from another work"
            predicates:
              - uri: "schema:isBasedOn"
              - uri: "wdt:P144"
                wikidata_label: "based on"
            domain: ["WRK.*"]
            range: ["WRK.*"]
            
          - code: "REL.WRK.TRN"
            label: "Translation Of"
            description: "Work is translation of another"
            predicates:
              - uri: "wdt:P629"
                wikidata_label: "edition or translation of"
            domain: ["WRK.TXT"]
            range: ["WRK.TXT"]
            
          - code: "REL.WRK.ADP"
            label: "Adaptation Of"
            description: "Work is adaptation of another"
            predicates:
              - uri: "wdt:P144"
                wikidata_label: "based on"
            domain: ["WRK.*"]
            range: ["WRK.*"]
            
          - code: "REL.WRK.REF"
            label: "References"
            description: "Work references another work"
            predicates:
              - uri: "dcterms:references"
              - uri: "schema:citation"
            domain: ["WRK.*"]
            range: ["WRK.*"]
            
          - code: "REL.WRK.PRT"
            label: "Part Of"
            description: "Work is part of larger work"
            predicates:
              - uri: "dcterms:isPartOf"
              - uri: "schema:isPartOf"
              - uri: "wdt:P361"
                wikidata_label: "part of"
            domain: ["WRK.*"]
            range: ["WRK.*"]
            
          - code: "REL.WRK.SER"
            label: "Part Of Series"
            description: "Work is part of a series"
            predicates:
              - uri: "schema:isPartOf"
              - uri: "wdt:P179"
                wikidata_label: "part of the series"
            domain: ["WRK.*"]
            range: ["WRK.SER"]

      # -------------------------------------------------------------------------
      # REL.CUS - Custodial/Ownership Relationships
      # -------------------------------------------------------------------------
      - code: "REL.CUS"
        label: "Custodial"
        description: |
          Relationships involving ownership, custody, and provenance of objects
          and collections. Critical for GLAM institutions.          
        ontology_mappings:
          - uri: "crm:P50_has_current_keeper"
            label: "has current keeper"
          - uri: "crm:P51_has_former_or_current_owner"
            label: "has former or current owner"
          - uri: "rico:hasOrHadHolder"
            label: "has or had holder"
        
        hyponyms:
          - code: "REL.CUS.OWN"
            label: "Owned By"
            description: "Object/collection owned by agent"
            predicates:
              - uri: "crm:P51_has_former_or_current_owner"
              - uri: "schema:ownedBy"
              - uri: "wdt:P127"
                wikidata_label: "owned by"
            domain: ["WRK.*", "WRK.OBJ"]
            range: ["AGT.PER", "GRP.*"]
            
          - code: "REL.CUS.KEP"
            label: "Kept By"
            description: "Object/collection in custody of agent"
            predicates:
              - uri: "crm:P50_has_current_keeper"
              - uri: "rico:hasOrHadHolder"
              - uri: "wdt:P195"
                wikidata_label: "collection"
            domain: ["WRK.*", "WRK.OBJ"]
            range: ["GRP.*"]
            
          - code: "REL.CUS.COL"
            label: "In Collection"
            description: "Item is in a collection"
            predicates:
              - uri: "edm:isShownAt"
              - uri: "wdt:P195"
                wikidata_label: "collection"
            domain: ["WRK.ITM", "WRK.OBJ"]
            range: ["GRP.*", "WRK.COL"]
            
          - code: "REL.CUS.PRV"
            label: "Previous Owner"
            description: "Object was previously owned by agent"
            predicates:
              - uri: "crm:P51_has_former_or_current_owner"
              - uri: "wdt:P127"
            domain: ["WRK.*", "WRK.OBJ"]
            range: ["AGT.PER", "GRP.*"]
            temporal: true
            
          - code: "REL.CUS.ACQ"
            label: "Acquired From"
            description: "Object acquired from agent"
            predicates:
              - uri: "crm:P24i_changed_ownership_through"
            domain: ["WRK.*", "WRK.OBJ"]
            range: ["AGT.PER", "GRP.*"]
            
          - code: "REL.CUS.DNT"
            label: "Donated By"
            description: "Object donated by agent"
            predicates:
              - uri: "wdt:P1028"
                wikidata_label: "donated by"
            domain: ["WRK.*", "WRK.OBJ"]
            range: ["AGT.PER", "GRP.*"]

      # -------------------------------------------------------------------------
      # REL.SUB - Subject/Topic Relationships
      # -------------------------------------------------------------------------
      - code: "REL.SUB"
        label: "Subject"
        description: |
          Relationships indicating what a work is about, depicts, or discusses.
          Critical for subject indexing and discovery.          
        ontology_mappings:
          - uri: "schema:about"
            label: "about"
          - uri: "dcterms:subject"
            label: "subject"
          - uri: "crm:P62_depicts"
            label: "depicts"
        
        hyponyms:
          - code: "REL.SUB.ABT"
            label: "About"
            description: "Work is about a topic/entity"
            predicates:
              - uri: "schema:about"
              - uri: "dcterms:subject"
              - uri: "wdt:P921"
                wikidata_label: "main subject"
            domain: ["WRK.*"]
            range: ["*"]
            
          - code: "REL.SUB.DEP"
            label: "Depicts"
            description: "Visual work depicts entity"
            predicates:
              - uri: "crm:P62_depicts"
              - uri: "wdt:P180"
                wikidata_label: "depicts"
            domain: ["WRK.VIS"]
            range: ["AGT.PER", "TOP.*", "WRK.OBJ"]
            
          - code: "REL.SUB.MNT"
            label: "Mentions"
            description: "Work mentions entity (not main subject)"
            predicates:
              - uri: "schema:mentions"
            domain: ["WRK.*"]
            range: ["*"]
            
          - code: "REL.SUB.SET"
            label: "Set In"
            description: "Narrative work set in place/time"
            predicates:
              - uri: "schema:contentLocation"
              - uri: "wdt:P840"
                wikidata_label: "narrative location"
            domain: ["WRK.TXT", "WRK.CIN", "WRK.PER"]
            range: ["TOP.*", "TMP.*"]
            
          - code: "REL.SUB.GNR"
            label: "Genre"
            description: "Work belongs to genre"
            predicates:
              - uri: "schema:genre"
              - uri: "wdt:P136"
                wikidata_label: "genre"
            domain: ["WRK.*"]
            range: ["WRK.GNR"]

      # -------------------------------------------------------------------------
      # REL.EVT - Event Relationships
      # -------------------------------------------------------------------------
      - code: "REL.EVT"
        label: "Event"
        description: |
          Relationships involving participation in or connection to events.
          Events are temporal entities with participants and locations.          
        ontology_mappings:
          - uri: "crm:P11_had_participant"
            label: "had participant"
          - uri: "crm:P7_took_place_at"
            label: "took place at"
          - uri: "schema:event"
            label: "event"
        
        hyponyms:
          - code: "REL.EVT.PAR"
            label: "Participated In"
            description: "Agent participated in event"
            predicates:
              - uri: "crm:P11i_participated_in"
              - uri: "wdt:P1344"
                wikidata_label: "participant in"
            domain: ["AGT.*"]
            range: ["EVT.*"]
            
          - code: "REL.EVT.ORG"
            label: "Organized"
            description: "Agent organized event"
            predicates:
              - uri: "schema:organizer"
              - uri: "wdt:P664"
                wikidata_label: "organizer"
            domain: ["AGT.*", "GRP.*"]
            range: ["EVT.*"]
            
          - code: "REL.EVT.LOC"
            label: "Event Location"
            description: "Event took place at location"
            predicates:
              - uri: "crm:P7_took_place_at"
              - uri: "schema:location"
              - uri: "wdt:P276"
                wikidata_label: "location"
            domain: ["EVT.*"]
            range: ["TOP.*"]
            
          - code: "REL.EVT.TIM"
            label: "Event Time"
            description: "Event occurred at time"
            predicates:
              - uri: "crm:P4_has_time-span"
              - uri: "schema:startDate"
              - uri: "wdt:P585"
                wikidata_label: "point in time"
            domain: ["EVT.*"]
            range: ["TMP.*"]
            
          - code: "REL.EVT.RSL"
            label: "Resulted In"
            description: "Event resulted in outcome"
            predicates:
              - uri: "prov:generated"
              - uri: "crm:P92_brought_into_existence"
            domain: ["EVT.*"]
            range: ["*"]

      # -------------------------------------------------------------------------
      # REL.IDT - Identity Relationships
      # -------------------------------------------------------------------------
      - code: "REL.IDT"
        label: "Identity"
        description: |
          Relationships expressing identity, equivalence, and representation.
          Distinct from coreference (same entity mentions) - these are
          ontological identity claims between different conceptualizations.          
        ontology_mappings:
          - uri: "owl:sameAs"
            label: "same as"
          - uri: "skos:exactMatch"
            label: "exact match"
          - uri: "skos:closeMatch"
            label: "close match"
        
        hyponyms:
          - code: "REL.IDT.SAM"
            label: "Same As"
            description: "Entity is identical to another (different URIs)"
            predicates:
              - uri: "owl:sameAs"
              - uri: "skos:exactMatch"
            domain: ["*"]
            range: ["*"]
            notes: |
              Use for linking to external authority files (VIAF, Wikidata, etc.)
              Distinct from coreference which links mentions within documents.              
            
          - code: "REL.IDT.REP"
            label: "Represents"
            description: "Entity represents another (e.g., portrait of person)"
            predicates:
              - uri: "crm:P138_represents"
            domain: ["WRK.VIS", "WRK.OBJ"]
            range: ["AGT.PER", "TOP.*"]
            
          - code: "REL.IDT.SIM"
            label: "Similar To"
            description: "Entity is similar to another (not identical)"
            predicates:
              - uri: "skos:closeMatch"
              - uri: "skos:relatedMatch"
            domain: ["*"]
            range: ["*"]
            
          - code: "REL.IDT.ALT"
            label: "Alternative Name For"
            description: "Appellation is alternative for entity"
            predicates:
              - uri: "skos:altLabel"
              - uri: "crm:P139_has_alternative_form"
            domain: ["APP.*"]
            range: ["*"]

      # -------------------------------------------------------------------------
      # REL.ROL - Role Relationships
      # -------------------------------------------------------------------------
      - code: "REL.ROL"
        label: "Role"
        description: |
          Relationships expressing roles that agents play in contexts.
          Temporally-bound role assertions.          
        ontology_mappings:
          - uri: "crm:P14.1_in_the_role_of"
            label: "in the role of"
          - uri: "schema:roleName"
            label: "role name"
        
        hyponyms:
          - code: "REL.ROL.HLD"
            label: "Held Role"
            description: "Agent held role/position"
            predicates:
              - uri: "schema:hasOccupation"
              - uri: "wdt:P39"
                wikidata_label: "position held"
            domain: ["AGT.PER"]
            range: ["ROL.*"]
            temporal: true
            
          - code: "REL.ROL.TTL"
            label: "Held Title"
            description: "Agent held title/rank"
            predicates:
              - uri: "wdt:P97"
                wikidata_label: "noble title"
              - uri: "wdt:P410"
                wikidata_label: "military rank"
            domain: ["AGT.PER"]
            range: ["ROL.TTL"]
            temporal: true
            
          - code: "REL.ROL.OCC"
            label: "Had Occupation"
            description: "Agent had occupation"
            predicates:
              - uri: "schema:hasOccupation"
              - uri: "wdt:P106"
                wikidata_label: "occupation"
            domain: ["AGT.PER"]
            range: ["ROL.OCC"]
            temporal: true
            
          - code: "REL.ROL.CTX"
            label: "Role Context"
            description: "Role was held in context of organization/event"
            predicates:
              - uri: "schema:worksFor"
              - uri: "crm:P11i_participated_in"
            domain: ["ROL.*"]
            range: ["GRP.*", "EVT.*"]

  # ---------------------------------------------------------------------------
  # RELATIONSHIP ANNOTATION SCHEMA
  # ---------------------------------------------------------------------------
  
  annotation_schema:
    description: |
      Schema for encoding relationship annotations as structured data.
      Supports both inline document annotations and standalone relationship
      assertions for knowledge graph construction.      
    
    required_fields:
      - field: "relationship_id"
        type: "string"
        format: "UUID or URI"
        description: "Unique identifier for this relationship annotation"
        
      - field: "relationship_type"
        type: "string"
        format: "REL.{HYPERNYM}.{HYPONYM}"
        description: "Relationship type code from taxonomy"
        examples:
          - "REL.CRE.AUT"
          - "REL.SOC.FAM.PAR"
          - "REL.WRK.DRV"
        
      - field: "subject"
        type: "object"
        description: "Source entity of the relationship"
        schema:
          entity_id: "string (reference to entity annotation)"
          entity_type: "string (entity type code)"
          span_text: "string (optional, surface form)"
          uri: "string (optional, external URI)"
          
      - field: "predicate"
        type: "object"
        description: "The relationship predicate"
        schema:
          uri: "string (ontology predicate URI)"
          label: "string (human-readable label)"
          direction: "enum (FORWARD, INVERSE)"
          
      - field: "object"
        type: "object"
        description: "Target entity of the relationship"
        schema:
          entity_id: "string (reference to entity annotation)"
          entity_type: "string (entity type code)"
          span_text: "string (optional, surface form)"
          uri: "string (optional, external URI)"
    
    optional_fields:
      - field: "temporal_scope"
        type: "object"
        description: "When the relationship held/holds"
        schema:
          start_date: "string (ISO 8601)"
          end_date: "string (ISO 8601 or null for ongoing)"
          temporal_modifier: "enum (CIRCA, BEFORE, AFTER, DURING)"
        examples:
          - start_date: "1632-01-01"
            end_date: "1669-10-04"
            temporal_modifier: null
          - start_date: "1890"
            end_date: null
            temporal_modifier: "CIRCA"
            
      - field: "spatial_scope"
        type: "object"
        description: "Where the relationship held/holds"
        schema:
          place_id: "string (reference to place entity)"
          place_name: "string"
          geo_uri: "string (GeoNames, Pleiades, etc.)"
          
      - field: "provenance"
        type: "object"
        description: "Source and confidence of relationship assertion"
        schema:
          source_document: "string (document ID)"
          source_span: "object (character offsets)"
          extraction_method: "enum (MANUAL, RULE_BASED, ML_EXTRACTED, EXTERNAL_KB)"
          extractor_id: "string (model or annotator ID)"
          extraction_date: "string (ISO 8601)"
          confidence: "float (0.0-1.0)"
          evidence_text: "string (supporting text)"
          
      - field: "qualifiers"
        type: "array"
        description: "Additional qualifications on the relationship"
        item_schema:
          qualifier_type: "string (e.g., 'role', 'capacity', 'manner')"
          qualifier_value: "string"
          qualifier_uri: "string (optional)"
        examples:
          - qualifier_type: "role"
            qualifier_value: "co-author"
          - qualifier_type: "capacity"
            qualifier_value: "as regent"
            
      - field: "negation"
        type: "boolean"
        default: false
        description: "True if relationship is explicitly negated"
        
      - field: "hypothetical"
        type: "boolean"
        default: false
        description: "True if relationship is hypothetical/uncertain"
        
      - field: "source_claim"
        type: "boolean"
        default: false
        description: "True if relationship is attributed to a source (not factual assertion)"
        
      - field: "attributed_to"
        type: "object"
        description: "If source_claim=true, who made the claim"
        schema:
          agent_id: "string"
          agent_name: "string"

  # ---------------------------------------------------------------------------
  # RELATIONSHIP EXTRACTION PATTERNS
  # ---------------------------------------------------------------------------
  
  extraction_patterns:
    description: |
      Linguistic patterns for extracting relationships from text.
      Patterns are language-specific and map to relationship types.      
    
    pattern_types:
      - type: "VERBAL"
        description: "Verb-mediated relationships"
        example: "Rembrandt painted The Night Watch"
        extraction_notes: |
          Parse verb as predicate, subject as source, object as target.
          Handle passive voice inversion.          
        
      - type: "NOMINAL"
        description: "Noun phrase relationships"
        example: "Rembrandt's painting The Night Watch"
        extraction_notes: |
          Possessive constructions imply creation/ownership.
          Appositive constructions may indicate identity.          
        
      - type: "PREPOSITIONAL"
        description: "Preposition-mediated relationships"
        example: "The Night Watch by Rembrandt"
        extraction_notes: |
          Prepositions indicate specific relationship types.
          "by" → creation, "in" → location, "of" → part-whole          
        
      - type: "COPULAR"
        description: "Is-a relationships"
        example: "Rembrandt was a painter"
        extraction_notes: |
          Copular verbs indicate typing, role, identity.
          Handle tense for temporal scoping.          
        
      - type: "COMPOUND"
        description: "Compound noun relationships"
        example: "the Rembrandt painting"
        extraction_notes: |
          Noun-noun compounds encode implicit relationships.
          Requires domain knowledge to disambiguate.          
    
    language_patterns:
      english:
        creation:
          - pattern: "{AGENT} (painted|wrote|composed|created|made|designed) {WORK}"
            relationship: "REL.CRE.*"
          - pattern: "{WORK} by {AGENT}"
            relationship: "REL.CRE.*"
          - pattern: "{AGENT}'s {WORK}"
            relationship: "REL.CRE.*"
            
        family:
          - pattern: "{PERSON} (son|daughter|child) of {PERSON}"
            relationship: "REL.SOC.FAM.CHD"
          - pattern: "{PERSON} (father|mother|parent) of {PERSON}"
            relationship: "REL.SOC.FAM.PAR"
          - pattern: "{PERSON} married {PERSON}"
            relationship: "REL.SOC.FAM.SPO"
            
        location:
          - pattern: "{ENTITY} in {PLACE}"
            relationship: "REL.SPA.LOC"
          - pattern: "{ENTITY} at {PLACE}"
            relationship: "REL.SPA.LOC"
          - pattern: "{ENTITY} from {PLACE}"
            relationship: "REL.SPA.ORG"
            
        temporal:
          - pattern: "{ENTITY} in {DATE}"
            relationship: "REL.TMP.DUR"
          - pattern: "{ENTITY} (during|throughout) {PERIOD}"
            relationship: "REL.TMP.DUR"
          - pattern: "{ENTITY} (before|prior to) {EVENT/DATE}"
            relationship: "REL.TMP.BEF"
            
      dutch:
        creation:
          - pattern: "{AGENT} (schilderde|schreef|componeerde|maakte) {WORK}"
            relationship: "REL.CRE.*"
          - pattern: "{WORK} van {AGENT}"
            relationship: "REL.CRE.*"
            
        family:
          - pattern: "{PERSON} (zoon|dochter|kind) van {PERSON}"
            relationship: "REL.SOC.FAM.CHD"
          - pattern: "{PERSON} (vader|moeder|ouder) van {PERSON}"
            relationship: "REL.SOC.FAM.PAR"
          - pattern: "{PERSON} trouwde met {PERSON}"
            relationship: "REL.SOC.FAM.SPO"
            
      german:
        creation:
          - pattern: "{AGENT} (malte|schrieb|komponierte|schuf) {WORK}"
            relationship: "REL.CRE.*"
          - pattern: "{WORK} von {AGENT}"
            relationship: "REL.CRE.*"
            
        family:
          - pattern: "{PERSON} (Sohn|Tochter|Kind) von {PERSON}"
            relationship: "REL.SOC.FAM.CHD"

  # ---------------------------------------------------------------------------
  # RELATIONSHIP REIFICATION
  # ---------------------------------------------------------------------------
  
  reification:
    description: |
      Mechanism for making statements about relationships themselves.
      Required for provenance, uncertainty, temporal qualification,
      and nested relationship structures.      
    
    approaches:
      - name: "RDF Reification"
        standard: "W3C RDF"
        description: "Classic quad-based reification using rdf:Statement"
        example: |
          _:stmt1 a rdf:Statement ;
            rdf:subject :Rembrandt ;
            rdf:predicate :created ;
            rdf:object :NightWatch ;
            :confidence 0.95 ;
            :source :Document123 .          
        limitations: "Verbose, query complexity"
        
      - name: "Named Graphs"
        standard: "W3C RDF 1.1"
        description: "Group triples in named graphs for provenance"
        example: |
          GRAPH :assertion123 {
            :Rembrandt :created :NightWatch .
          }
          :assertion123 :confidence 0.95 ;
                        :source :Document123 .          
        advantages: "Cleaner queries, SPARQL support"
        
      - name: "Wikibase Qualifiers"
        standard: "Wikidata model"
        description: "Qualifiers attached to statement nodes"
        example: |
          wd:Q5598 p:P170 [
            ps:P170 wd:Q5582 ;
            pq:P580 "1642"^^xsd:gYear ;
            prov:wasDerivedFrom [ pr:P248 wd:Q12345 ]
          ] .          
        advantages: "Rich qualification, widespread adoption"
        
      - name: "RDF-star"
        standard: "W3C RDF-star (emerging)"
        description: "Quoted triples as subjects/objects"
        example: |
          << :Rembrandt :created :NightWatch >> :confidence 0.95 .          
        advantages: "Concise, intuitive"
        limitations: "Limited tooling support (as of 2025)"
    
    recommended_approach: "Named Graphs"
    rationale: |
      Named graphs provide the best balance of expressivity, query support,
      and tooling maturity. Each relationship annotation becomes a named
      graph containing the core triple, with provenance attached to the
      graph identifier.      
    
    schema:
      graph_naming: |
        Graph URIs follow pattern:
        {document_uri}#rel-{relationship_id}
        
        Example:
        https://example.org/doc/123#rel-a1b2c3d4        
      
      required_metadata:
        - "prov:generatedAtTime"
        - "prov:wasAttributedTo"
        - "oa:hasBody (the relationship triple)"
        - "oa:hasTarget (source document span)"

  # ---------------------------------------------------------------------------
  # RELATIONSHIP VALIDATION
  # ---------------------------------------------------------------------------
  
  validation:
    description: |
      Rules for validating relationship annotations. Ensures consistency
      with entity types, ontology constraints, and domain rules.      
    
    type_constraints:
      description: "Validate domain/range compatibility"
      rules:
        - rule: "Subject entity type must be in relationship domain"
          error_code: "REL_DOMAIN_VIOLATION"
          example: |
            INVALID: REL.CRE.AUT with subject type TOP.BLT (building can't author)
            VALID: REL.CRE.AUT with subject type AGT.PER (person can author)
                        
        - rule: "Object entity type must be in relationship range"
          error_code: "REL_RANGE_VIOLATION"
          example: |
            INVALID: REL.CRE.AUT with object type AGT.PER (can't author a person)
            VALID: REL.CRE.AUT with object type WRK.TXT (can author a text)            
    
    logical_constraints:
      description: "Validate logical consistency"
      rules:
        - rule: "Symmetric relationships must have inverse annotation"
          applies_to: ["REL.SOC.FAM.SIB", "REL.SOC.FAM.SPO", "REL.SPA.ADJ"]
          
        - rule: "Asymmetric relationships cannot have same subject and object"
          applies_to: ["REL.SOC.FAM.PAR", "REL.ORG.PAR"]
          
        - rule: "Temporal relationships must have valid date ordering"
          applies_to: ["REL.TMP.BEF", "REL.TMP.AFT"]
          example: |
            INVALID: Event A REL.TMP.BEF Event B where A.date > B.date            
    
    cardinality_constraints:
      description: "Validate multiplicity rules"
      rules:
        - rule: "Person can have at most 2 biological parents"
          relationship: "REL.SOC.FAM.PAR"
          max_cardinality: 2
          
        - rule: "Work has exactly one original creator (primary)"
          relationship: "REL.CRE.*"
          notes: "Secondary creators use qualifiers"

  # ---------------------------------------------------------------------------
  # EXAMPLES
  # ---------------------------------------------------------------------------
  
  examples:
    - name: "Creation relationship"
      source_text: "The Night Watch was painted by Rembrandt in 1642."
      annotation:
        relationship_id: "rel-001"
        relationship_type: "REL.CRE.ART"
        subject:
          entity_id: "ent-rembrandt"
          entity_type: "AGT.PER"
          span_text: "Rembrandt"
          uri: "http://viaf.org/viaf/64013650"
        predicate:
          uri: "http://www.cidoc-crm.org/cidoc-crm/P14_carried_out_by"
          label: "carried out by"
          direction: "INVERSE"
        object:
          entity_id: "ent-nightwatch"
          entity_type: "WRK.VIS"
          span_text: "The Night Watch"
          uri: "http://www.wikidata.org/entity/Q219831"
        temporal_scope:
          start_date: "1642"
          end_date: "1642"
        provenance:
          source_document: "doc-123"
          source_span:
            start: 0
            end: 47
          extraction_method: "RULE_BASED"
          confidence: 0.95
          evidence_text: "The Night Watch was painted by Rembrandt in 1642."

    - name: "Family relationship with temporal scope"
      source_text: "Willem-Alexander, son of Queen Beatrix, became king in 2013."
      annotation:
        relationship_id: "rel-002"
        relationship_type: "REL.SOC.FAM.CHD"
        subject:
          entity_id: "ent-willem"
          entity_type: "AGT.PER"
          span_text: "Willem-Alexander"
        predicate:
          uri: "http://purl.org/vocab/relationship/childOf"
          label: "child of"
          direction: "FORWARD"
        object:
          entity_id: "ent-beatrix"
          entity_type: "AGT.PER"
          span_text: "Queen Beatrix"
        temporal_scope:
          start_date: "1967-04-27"
          end_date: null
        provenance:
          extraction_method: "MANUAL"
          confidence: 1.0
          
    - name: "Custodial relationship"
      source_text: "The Rijksmuseum houses over 8,000 objects from Rembrandt's era."
      annotation:
        relationship_id: "rel-003"
        relationship_type: "REL.CUS.KEP"
        subject:
          entity_id: "ent-rijks-collection"
          entity_type: "WRK.COL"
          span_text: "8,000 objects"
        predicate:
          uri: "http://www.cidoc-crm.org/cidoc-crm/P50_has_current_keeper"
          label: "has current keeper"
          direction: "FORWARD"
        object:
          entity_id: "ent-rijksmuseum"
          entity_type: "GRP.ORG"
          span_text: "The Rijksmuseum"
          uri: "http://www.wikidata.org/entity/Q190804"
        provenance:
          extraction_method: "ML_EXTRACTED"
          confidence: 0.88

    - name: "Work derivation (FRBR)"
      source_text: "The 2019 film 'Little Women' is based on Louisa May Alcott's 1868 novel."
      annotation:
        relationship_id: "rel-004"
        relationship_type: "REL.WRK.ADP"
        subject:
          entity_id: "ent-lw-film"
          entity_type: "WRK.CIN"
          span_text: "The 2019 film 'Little Women'"
        predicate:
          uri: "http://schema.org/isBasedOn"
          label: "is based on"
          direction: "FORWARD"
        object:
          entity_id: "ent-lw-novel"
          entity_type: "WRK.TXT"
          span_text: "Louisa May Alcott's 1868 novel"
        provenance:
          extraction_method: "RULE_BASED"
          confidence: 0.98

    - name: "Negated relationship"
      source_text: "Despite popular belief, Mozart did not poison Salieri."
      annotation:
        relationship_id: "rel-005"
        relationship_type: "REL.EVT.PAR"
        subject:
          entity_id: "ent-mozart"
          entity_type: "AGT.PER"
          span_text: "Mozart"
        predicate:
          uri: "http://www.cidoc-crm.org/cidoc-crm/P11i_participated_in"
          label: "participated in"
          direction: "FORWARD"
        object:
          entity_id: "ent-poisoning-event"
          entity_type: "EVT.ACT"
          span_text: "poison Salieri"
        negation: true
        provenance:
          extraction_method: "MANUAL"
          confidence: 1.0
          evidence_text: "Despite popular belief, Mozart did not poison Salieri."

# =============================================================================
# SECTION 17: COREFERENCE RESOLUTION
# =============================================================================
# 
# This section defines rules for identifying when multiple mentions in text
# refer to the same real-world entity. Distinct from relationship annotations
# which link DIFFERENT entities.
#
# Coreference is fundamental to:
#   - Entity consolidation (grouping mentions into entity clusters)
#   - Knowledge graph construction (single node per entity)
#   - Information extraction (aggregating facts about an entity)
#   - Document understanding (tracking entities through narrative)
#
# Ontological foundations:
#   - OWL: owl:sameAs for identity assertions
#   - SKOS: skos:exactMatch for vocabulary alignment
#   - NIF: nif:referenceContext for text anchoring
#   - CIDOC-CRM: E42 Identifier for naming
#   - TEI: <rs> (referring string) with @ref and @corresp
#
# Linguistic foundations:
#   - Centering Theory (Grosz, Joshi, Weinstein)
#   - Binding Theory (Chomsky)
#   - Discourse Representation Theory (Kamp)
#   - ACE/OntoNotes coreference guidelines
# =============================================================================

coreference_resolution:
  
  description: |
    Coreference resolution identifies when two or more text spans refer to
    the same entity in the real world (or fictional world for narratives).
    
    KEY DISTINCTION:
    - COREFERENCE: "Rembrandt" and "the Dutch master" → SAME entity
    - RELATIONSHIP: "Rembrandt" and "The Night Watch" → DIFFERENT entities linked by creation
    
    Coreference creates ENTITY CLUSTERS where each cluster represents one
    real-world entity and contains all textual mentions of that entity.    
  
  namespaces:
    nif: "http://persistence.uni-leipzig.org/nlp2rdf/ontologies/nif-core#"
    itsrdf: "http://www.w3.org/2005/11/its/rdf#"
    conll: "http://ufal.mff.cuni.cz/conll2009-st/task-description.html#"
    ace: "https://www.ldc.upenn.edu/collaborations/past-projects/ace"
    ontonotes: "https://catalog.ldc.upenn.edu/LDC2013T19"
    owl: "http://www.w3.org/2002/07/owl#"
    skos: "http://www.w3.org/2004/02/skos/core#"
    crm: "http://www.cidoc-crm.org/cidoc-crm/"
    tei: "http://www.tei-c.org/ns/1.0"
    prov: "http://www.w3.org/ns/prov#"

  # ---------------------------------------------------------------------------
  # COREFERENCE TYPES
  # ---------------------------------------------------------------------------
  
  coreference_types:
    description: |
      Classification of coreference phenomena by linguistic mechanism.
      Each type has distinct annotation requirements and resolution strategies.      
    
    types:
      # -------------------------------------------------------------------------
      # COREF.PRO - Pronominal Reference
      # -------------------------------------------------------------------------
      - code: "COREF.PRO"
        label: "Pronominal"
        description: |
          Pronouns referring to antecedent noun phrases. The most common
          coreference type. Requires grammatical agreement (person, number,
          gender) and discourse salience.          
        
        subtypes:
          - code: "COREF.PRO.PER"
            label: "Personal Pronoun"
            description: "he, she, they, it, him, her, them"
            examples:
              - text: "Rembrandt painted portraits. He was a master of light."
                antecedent: "Rembrandt"
                anaphor: "He"
                
          - code: "COREF.PRO.POS"
            label: "Possessive Pronoun"
            description: "his, her, their, its"
            examples:
              - text: "The museum expanded its collection in 2020."
                antecedent: "The museum"
                anaphor: "its"
                
          - code: "COREF.PRO.REF"
            label: "Reflexive Pronoun"
            description: "himself, herself, themselves, itself"
            examples:
              - text: "The artist painted himself in the corner."
                antecedent: "The artist"
                anaphor: "himself"
            binding_constraint: "Must be in same clause as antecedent"
            
          - code: "COREF.PRO.REC"
            label: "Reciprocal Pronoun"
            description: "each other, one another"
            examples:
              - text: "The artists influenced each other."
                antecedent: "The artists"
                anaphor: "each other"
                
          - code: "COREF.PRO.DEM"
            label: "Demonstrative Pronoun"
            description: "this, that, these, those (standalone)"
            examples:
              - text: "The painting sold for millions. This surprised critics."
                antecedent: "The painting sold for millions"
                anaphor: "This"
                notes: "Can refer to propositions/events, not just entities"
                
          - code: "COREF.PRO.REL"
            label: "Relative Pronoun"
            description: "who, which, that, whose"
            examples:
              - text: "Vermeer, who lived in Delft, painted domestic scenes."
                antecedent: "Vermeer"
                anaphor: "who"
            binding_constraint: "Immediate syntactic dependency on antecedent"

      # -------------------------------------------------------------------------
      # COREF.NOM - Nominal Reference
      # -------------------------------------------------------------------------
      - code: "COREF.NOM"
        label: "Nominal"
        description: |
          Full noun phrases referring to previously mentioned entities.
          Includes definite descriptions, epithets, and proper noun variants.          
        
        subtypes:
          - code: "COREF.NOM.DEF"
            label: "Definite Description"
            description: "NP with definite article referring to known entity"
            examples:
              - text: "Rembrandt was born in Leiden. The painter moved to Amsterdam."
                antecedent: "Rembrandt"
                anaphor: "The painter"
              - text: "The Rijksmuseum reopened in 2013. The museum attracted millions."
                antecedent: "The Rijksmuseum"
                anaphor: "The museum"
            annotation_notes: |
              Definite descriptions may be:
              - Coreferent (same entity)
              - Bridging (related entity, see COREF.BRD)
              Disambiguation requires world knowledge.
                              
          - code: "COREF.NOM.EPI"
            label: "Epithet"
            description: "Descriptive phrase used as name substitute"
            examples:
              - text: "Vermeer captured light beautifully. The Sphinx of Delft remains mysterious."
                antecedent: "Vermeer"
                anaphor: "The Sphinx of Delft"
              - text: "Napoleon dominated Europe. The Little Corporal met his end at Waterloo."
                antecedent: "Napoleon"
                anaphor: "The Little Corporal"
            annotation_notes: |
              Epithets are often culturally-specific and may require
              domain knowledge to resolve. Flag for verification when
              confidence is low.
                              
          - code: "COREF.NOM.NAM"
            label: "Name Variant"
            description: "Different forms of the same proper name"
            examples:
              - text: "Rembrandt Harmenszoon van Rijn was born in 1606. Rembrandt is known for self-portraits."
                antecedent: "Rembrandt Harmenszoon van Rijn"
                anaphor: "Rembrandt"
              - text: "The Metropolitan Museum of Art opened in 1872. The Met now has 2 million works."
                antecedent: "The Metropolitan Museum of Art"
                anaphor: "The Met"
            annotation_notes: |
              Name variants include:
              - Abbreviations (Metropolitan → Met)
              - Nicknames (William → Bill)
              - Formal/informal (Dr. Smith → John)
              - Transliterations (Tchaikovsky / Čajkovskij)
                              
          - code: "COREF.NOM.REP"
            label: "Repetition"
            description: "Exact or near-exact repetition of noun phrase"
            examples:
              - text: "The Night Watch hangs in the Rijksmuseum. The Night Watch was restored in 2019."
                antecedent: "The Night Watch"
                anaphor: "The Night Watch"
            annotation_notes: |
              Repetitions are high-confidence coreferences but verify
              for potential ambiguity (e.g., multiple paintings with
              similar names).              

      # -------------------------------------------------------------------------
      # COREF.ZER - Zero Anaphora
      # -------------------------------------------------------------------------
      - code: "COREF.ZER"
        label: "Zero Anaphora"
        description: |
          Implicit reference where the anaphor is not realized in surface text.
          Common in pro-drop languages (Spanish, Japanese, Chinese) and in
          coordinate structures in English.          
        
        subtypes:
          - code: "COREF.ZER.PRO"
            label: "Pro-drop"
            description: "Omitted subject pronoun (pro-drop languages)"
            examples:
              - text: "El artista pintó retratos. [Ø] Era un maestro."
                language: "Spanish"
                antecedent: "El artista"
                anaphor: "[implicit subject]"
              - text: "画家が肖像画を描いた。[Ø]光の達人だった。"
                language: "Japanese"
                antecedent: "画家"
                anaphor: "[implicit subject]"
                
          - code: "COREF.ZER.GAP"
            label: "Gapping"
            description: "Ellipsis in coordinate structures"
            examples:
              - text: "Vermeer painted interiors and Rembrandt [Ø] portraits."
                antecedent: "painted"
                anaphor: "[elided verb]"
                notes: "This is VP ellipsis, not entity coreference"
                
          - code: "COREF.ZER.SLU"
            label: "Sluicing"
            description: "Ellipsis in embedded questions"
            examples:
              - text: "Someone stole the painting, but we don't know who [Ø]."
                antecedent: "stole the painting"
                anaphor: "[elided VP]"

      # -------------------------------------------------------------------------
      # COREF.CAT - Cataphora
      # -------------------------------------------------------------------------
      - code: "COREF.CAT"
        label: "Cataphora"
        description: |
          Forward-pointing reference where the pronoun PRECEDES its referent.
          Less common than anaphora; creates suspense or emphasis.          
        
        examples:
          - text: "When he arrived in Amsterdam, Rembrandt was only 25."
            cataphor: "he"
            postcedent: "Rembrandt"
          - text: "Although it was controversial, The Night Watch became Rembrandt's most famous work."
            cataphor: "it"
            postcedent: "The Night Watch"
            
        constraints:
          - "Cataphor typically in subordinate clause"
          - "Postcedent in main clause"
          - "Binding theory: c-command constraints apply"

      # -------------------------------------------------------------------------
      # COREF.BRD - Bridging Reference
      # -------------------------------------------------------------------------
      - code: "COREF.BRD"
        label: "Bridging"
        description: |
          Reference to an entity not explicitly mentioned but inferrable from
          context via a relation to an mentioned entity. NOT strict coreference
          but essential for discourse coherence.          
        
        subtypes:
          - code: "COREF.BRD.PRT"
            label: "Part-Whole Bridging"
            description: "Reference to a part of a mentioned whole"
            examples:
              - text: "The museum was renovated. The roof was replaced."
                anchor: "The museum"
                bridged: "The roof"
                relation: "part-of"
              - text: "I bought a book. The first chapter was fascinating."
                anchor: "a book"
                bridged: "The first chapter"
                relation: "part-of"
                
          - code: "COREF.BRD.SET"
            label: "Set-Member Bridging"
            description: "Reference to a member of a mentioned set"
            examples:
              - text: "The paintings arrived. The largest was a Rubens."
                anchor: "The paintings"
                bridged: "The largest"
                relation: "member-of"
                
          - code: "COREF.BRD.EVT"
            label: "Event Bridging"
            description: "Reference to participant in mentioned event"
            examples:
              - text: "The auction concluded. The winning bidder remained anonymous."
                anchor: "The auction"
                bridged: "The winning bidder"
                relation: "participant-in"
                
          - code: "COREF.BRD.ATR"
            label: "Attribute Bridging"
            description: "Reference to attribute of mentioned entity"
            examples:
              - text: "The painting is impressive. The size overwhelms visitors."
                anchor: "The painting"
                bridged: "The size"
                relation: "attribute-of"
                
          - code: "COREF.BRD.PRD"
            label: "Producer/Product Bridging"
            description: "Reference via production relationship"
            examples:
              - text: "Rembrandt was prolific. The paintings number over 300."
                anchor: "Rembrandt"
                bridged: "The paintings"
                relation: "produced-by"
                
        annotation_notes: |
          Bridging is DISTINCT from coreference:
          - Coreference: same entity, different mentions
          - Bridging: different entities, related by inference
          
          Bridging annotations should specify the bridging RELATION
          and the ANCHOR entity that enables the inference.          

      # -------------------------------------------------------------------------
      # COREF.SPL - Split Antecedent
      # -------------------------------------------------------------------------
      - code: "COREF.SPL"
        label: "Split Antecedent"
        description: |
          Plural pronoun referring to multiple previously mentioned entities
          that together form the antecedent.          
        
        examples:
          - text: "Rembrandt met Saskia in 1633. They married the following year."
            antecedents: ["Rembrandt", "Saskia"]
            anaphor: "They"
          - text: "The curator and the director disagreed. They eventually compromised."
            antecedents: ["The curator", "the director"]
            anaphor: "They"
            
        annotation_notes: |
          Split antecedents create a GROUP entity from individual mentions.
          The annotation should:
          1. Link the anaphor to each antecedent mention
          2. Create an implicit group entity
          3. Record that this is split-antecedent resolution          

      # -------------------------------------------------------------------------
      # COREF.GEN - Generic Reference
      # -------------------------------------------------------------------------
      - code: "COREF.GEN"
        label: "Generic"
        description: |
          Reference to generic/kind-level entities rather than specific
          individuals. Requires special handling as generic mentions don't
          refer to token-level entities.          
        
        subtypes:
          - code: "COREF.GEN.KND"
            label: "Kind Reference"
            description: "Reference to a kind or class"
            examples:
              - text: "The dodo is extinct. It was native to Mauritius."
                kind: "dodo (species)"
                notes: "Both mentions refer to the KIND, not an individual"
                
          - code: "COREF.GEN.ARB"
            label: "Arbitrary Reference"
            description: "Reference to arbitrary/generic individual"
            examples:
              - text: "A museum curator must be knowledgeable. They work long hours."
                generic_type: "arbitrary curator"
                notes: "'They' refers to any curator, not a specific one"
                
        annotation_notes: |
          Generic coreference is often EXCLUDED from mention-level
          coreference annotation (per ACE/OntoNotes guidelines) but
          may be relevant for knowledge base population.
          
          Flag generic mentions with is_generic=true.          

      # -------------------------------------------------------------------------
      # COREF.APP - Appositive
      # -------------------------------------------------------------------------
      - code: "COREF.APP"
        label: "Appositive"
        description: |
          Noun phrase immediately following another NP that renames or
          describes the same entity. High-confidence coreference.          
        
        subtypes:
          - code: "COREF.APP.TIT"
            label: "Title/Role Appositive"
            examples:
              - text: "Rembrandt, the Dutch painter, was born in 1606."
                head: "Rembrandt"
                appositive: "the Dutch painter"
                
          - code: "COREF.APP.ALT"
            label: "Alternative Name Appositive"
            examples:
              - text: "The Night Watch, or De Nachtwacht, hangs in Amsterdam."
                head: "The Night Watch"
                appositive: "De Nachtwacht"
                
          - code: "COREF.APP.REL"
            label: "Relationship Appositive"
            examples:
              - text: "Saskia, Rembrandt's wife, died in 1642."
                head: "Saskia"
                appositive: "Rembrandt's wife"
                
        syntactic_pattern: "NP1, NP2, ..."
        annotation_notes: |
          Appositives provide HIGH CONFIDENCE coreference signals.
          They also often provide entity TYPE information
          (e.g., "the Dutch painter" → type=AGT.PER, occupation=painter)          

  # ---------------------------------------------------------------------------
  # COREFERENCE ANNOTATION SCHEMA
  # ---------------------------------------------------------------------------
  
  annotation_schema:
    description: |
      Schema for encoding coreference annotations. Supports both mention-pair
      and entity-cluster representations.      
    
    mention_annotation:
      description: "Annotation for individual mentions"
      required_fields:
        - field: "mention_id"
          type: "string"
          format: "UUID or sequential ID"
          description: "Unique identifier for this mention"
          
        - field: "span"
          type: "object"
          schema:
            start: "integer (character offset)"
            end: "integer (character offset)"
            text: "string (surface form)"
          description: "Text span of the mention"
          
        - field: "entity_type"
          type: "string"
          description: "Entity type code (AGT.PER, GRP.ORG, etc.)"
          
      optional_fields:
        - field: "head_span"
          type: "object"
          schema:
            start: "integer"
            end: "integer"
            text: "string"
          description: "Syntactic head of the mention (for complex NPs)"
          
        - field: "mention_type"
          type: "enum"
          values:
            - "PROPER"      # Proper noun (Rembrandt, Amsterdam)
            - "NOMINAL"     # Common noun phrase (the painter, a museum)
            - "PRONOMINAL"  # Pronoun (he, she, it, they)
            - "ZERO"        # Implicit/zero anaphor
          description: "Morphosyntactic type of mention"
          
        - field: "is_generic"
          type: "boolean"
          default: false
          description: "True if mention refers to a kind, not individual"
          
        - field: "grammatical_role"
          type: "enum"
          values: ["SUBJECT", "OBJECT", "OBLIQUE", "POSSESSIVE", "APPOSITIVE"]
          description: "Grammatical function in sentence"
          
        - field: "number"
          type: "enum"
          values: ["SINGULAR", "PLURAL", "UNKNOWN"]
          description: "Grammatical number"
          
        - field: "gender"
          type: "enum"
          values: ["MASCULINE", "FEMININE", "NEUTER", "UNKNOWN"]
          description: "Grammatical gender (language-dependent)"
          
        - field: "animacy"
          type: "enum"
          values: ["ANIMATE", "INANIMATE", "UNKNOWN"]
          description: "Animacy feature"
    
    cluster_annotation:
      description: |
        Entity cluster grouping all mentions of the same entity.
        The primary output of coreference resolution.        
      
      required_fields:
        - field: "cluster_id"
          type: "string"
          description: "Unique identifier for the entity cluster"
          
        - field: "mentions"
          type: "array"
          item_type: "mention_id (reference)"
          description: "All mentions belonging to this cluster"
          
        - field: "entity_type"
          type: "string"
          description: "Entity type of the cluster (most specific common type)"
          
      optional_fields:
        - field: "canonical_mention"
          type: "mention_id"
          description: "The 'best' mention for display (usually first proper noun)"
          
        - field: "external_uri"
          type: "string"
          format: "URI"
          description: "Link to external knowledge base (Wikidata, VIAF, etc.)"
          
        - field: "confidence"
          type: "float"
          range: [0.0, 1.0]
          description: "Confidence in cluster correctness"
          
        - field: "provenance"
          type: "object"
          schema:
            resolution_method: "enum (MANUAL, RULE_BASED, ML_NEURAL, HYBRID)"
            resolver_id: "string"
            resolution_date: "ISO 8601"
          description: "How the cluster was formed"
    
    link_annotation:
      description: |
        Explicit pairwise coreference links (alternative to cluster-first).
        Useful for bridging and split antecedents.        
      
      required_fields:
        - field: "link_id"
          type: "string"
          
        - field: "source_mention"
          type: "mention_id"
          description: "The anaphor (referring expression)"
          
        - field: "target_mention"
          type: "mention_id"
          description: "The antecedent (referred-to expression)"
          
        - field: "link_type"
          type: "string"
          format: "COREF.{TYPE}.{SUBTYPE}"
          description: "Coreference type from taxonomy"
          
      optional_fields:
        - field: "bridging_relation"
          type: "string"
          description: "For COREF.BRD, the semantic relation"
          values: ["part-of", "member-of", "attribute-of", "participant-in", "produced-by"]
          
        - field: "confidence"
          type: "float"

  # ---------------------------------------------------------------------------
  # RESOLUTION STRATEGIES
  # ---------------------------------------------------------------------------
  
  resolution_strategies:
    description: |
      Strategies for resolving coreference, from rule-based to neural approaches.      
    
    rule_based:
      description: "Deterministic rules based on linguistic constraints"
      
      rules:
        - name: "Exact String Match"
          description: "Identical strings likely corefer"
          confidence: 0.95
          exceptions:
            - "Common nouns in different contexts"
            - "Homonymous proper nouns (multiple people named 'John Smith')"
          
        - name: "Acronym Expansion"
          description: "Acronym matches expanded form"
          pattern: "Match 'BBC' with 'British Broadcasting Corporation'"
          confidence: 0.90
          
        - name: "Appositive"
          description: "NP in apposition corefers with head"
          pattern: "NP1, NP2 → NP1 = NP2"
          confidence: 0.98
          
        - name: "Predicate Nominative"
          description: "Subject and predicate nominative corefer"
          pattern: "X is Y → X = Y (when Y is identity statement)"
          confidence: 0.85
          
        - name: "Reflexive Binding"
          description: "Reflexive binds to local subject"
          pattern: "X ... himself/herself → reflexive = X"
          confidence: 0.99
          constraint: "Same clause"
          
        - name: "Relative Pronoun"
          description: "Relative pronoun binds to immediately preceding NP"
          pattern: "NP who/which ... → relative = NP"
          confidence: 0.99
          
        - name: "Gender Agreement"
          description: "Pronoun must match antecedent gender"
          constraint: "FILTER: he → masculine, she → feminine, it → neuter"
          
        - name: "Number Agreement"
          description: "Pronoun must match antecedent number"
          constraint: "FILTER: they → plural, he/she/it → singular"
          
        - name: "Recency Preference"
          description: "More recent mentions preferred as antecedents"
          notes: "All else equal, prefer closer antecedent"
          
        - name: "Grammatical Role Preference"
          description: "Subject > Object > Oblique for antecedent"
          notes: "Per Centering Theory"
    
    statistical:
      description: "Machine learning approaches to coreference"
      
      approaches:
        - name: "Mention-Pair Model"
          description: "Binary classification for mention pairs"
          features:
            - "String matching features"
            - "Distance features (sentence, mention)"
            - "Grammatical features"
            - "Semantic features (WordNet, embeddings)"
          limitations: "Doesn't model transitivity well"
          
        - name: "Mention-Ranking Model"
          description: "Rank antecedent candidates for each mention"
          advantages: "Handles competition between candidates"
          
        - name: "Entity-Centric Model"
          description: "Model clusters directly, not pairs"
          advantages: "Better global coherence"
          
        - name: "End-to-End Neural"
          description: "Jointly learn mention detection and coreference"
          sota_models:
            - "Lee et al. (2017) - span representations"
            - "Joshi et al. (2020) - SpanBERT-based"
            - "Kirstain et al. (2021) - s2e model"
          advantages: "No pipeline errors, learns complex patterns"
          limitations: "Requires large training data, less interpretable"

  # ---------------------------------------------------------------------------
  # DOMAIN-SPECIFIC CONSIDERATIONS
  # ---------------------------------------------------------------------------
  
  domain_considerations:
    
    heritage_domain:
      description: |
        Special considerations for cultural heritage documents
        (archival finding aids, museum catalogs, art history texts).        
      
      challenges:
        - challenge: "Historical Name Variants"
          description: "Same person may have multiple historical spellings"
          examples:
            - "Rembrandt / Rembrant / Rembrand"
            - "Titian / Tiziano Vecelli / Tiziano Vecellio"
          strategy: "Maintain authority file linkage; use VIAF/ULAN"
          
        - challenge: "Institutional Name Changes"
          description: "Organizations change names over time"
          examples:
            - "Mauritshuis / Koninklijk Kabinet van Schilderijen"
            - "The Tate / Tate Gallery / Tate Britain"
          strategy: "Track temporal validity of names; link to authority"
          
        - challenge: "Anonymous/Unknown Creators"
          description: "Attributions like 'Circle of Rembrandt', 'Follower of'"
          examples:
            - "Follower of Vermeer"
            - "Workshop of Rubens"
            - "Anonymous (Dutch, 17th century)"
          strategy: |
            Create placeholder entities for attribution groups.
            Do NOT corefer with named artists.
                        
        - challenge: "Work Titles"
          description: "Artworks may have multiple titles"
          examples:
            - "The Night Watch / De Nachtwacht / Militia Company of District II"
            - "Mona Lisa / La Gioconda / Portrait of Lisa Gherardini"
          strategy: "Canonical title in preferred language; track variants"
          
        - challenge: "Archival Hierarchies"
          description: "Nested references in finding aids"
          examples:
            - "The collection... this series... the folder..."
          strategy: "Model archival hierarchy; bridging for containment"
          
      entity_type_specifics:
        - type: "AGT.PER (Persons)"
          notes: |
            - Artists often referred to by last name only after first mention
            - Titles/honorifics may vary (van Gogh / Mr. van Gogh / Vincent)
            - Biographical facts help disambiguation
                        
        - type: "GRP.ORG (Organizations)"
          notes: |
            - Museums, archives may merge/split
            - Acronyms common (MoMA, V&A, BnF)
            - Departmental references may need bridging
                        
        - type: "WRK.* (Works)"
          notes: |
            - Works may be referred to by title, artist, or description
            - "The painting" may need context for resolution
            - Distinguish work vs. physical object (FRBR)
                        
        - type: "TOP.* (Places)"
          notes: |
            - Historical place names may differ from modern
            - "Delft" may be city or adjective ("Delft painter")
            - Nested places (room in museum in city)            
    
    multilingual:
      description: "Considerations for multilingual documents"
      
      challenges:
        - challenge: "Cross-lingual Coreference"
          description: "Same entity mentioned in different languages"
          examples:
            - "Amsterdam (EN) / Amsterdam (NL) / Ámsterdam (ES)"
            - "The Night Watch / De Nachtwacht"
          strategy: "Normalize via external authority (Wikidata)"
          
        - challenge: "Translation Variants"
          description: "Translated names may not be exact"
          examples:
            - "Rijksmuseum / State Museum / Musée national"
          strategy: "Fuzzy matching with translation knowledge"
          
        - challenge: "Pro-drop Languages"
          description: "Subject pronouns omitted in Spanish, Japanese, etc."
          strategy: "Use zero anaphora annotation; infer from verb agreement"

  # ---------------------------------------------------------------------------
  # ANNOTATION GUIDELINES
  # ---------------------------------------------------------------------------
  
  annotation_guidelines:
    
    markable_entities:
      description: "What counts as a markable mention"
      include:
        - "Named entities (proper nouns)"
        - "Definite noun phrases with specific referent"
        - "Pronouns with recoverable antecedent"
        - "Demonstrative NPs referring to specific entities"
        - "Possessive NPs (his, their, the museum's)"
      exclude:
        - "Expletive pronouns (It is raining, There are problems)"
        - "Idioms where reference is not compositional"
        - "Generic/kind-level references (optional exclusion)"
        - "Predicative NPs (He is a painter - 'a painter' is not markable)"
        - "Bound pronouns in quantified expressions (Every artist sells their work)"
    
    span_conventions:
      description: "How to delimit mention spans"
      rules:
        - rule: "Maximal NP Span"
          description: "Include full NP including determiners and modifiers"
          example: "[The famous Dutch painter Rembrandt] → entire span"
          
        - rule: "Nested Mentions"
          description: "Mark nested mentions separately"
          example: "[Rembrandt]'s [wife] → two mentions, linked"
          
        - rule: "Conjoined NPs"
          description: "Mark entire conjunction and individual conjuncts"
          example: "[[Rembrandt] and [Vermeer]] both painted → 3 mentions"
          
        - rule: "Head Annotation"
          description: "Additionally mark syntactic head within span"
          example: "[The famous painter (head=painter)]"
    
    edge_cases:
      - case: "Singleton Entities"
        description: "Entities mentioned only once"
        guideline: "Include in annotation; mark as singleton cluster"
        
      - case: "Discourse Deixis"
        description: "Reference to propositions/events, not entities"
        example: "'John left. That surprised me.' - 'That' refers to event"
        guideline: "Annotate with special type COREF.PRO.DEM; link to event span"
        
      - case: "Metonymy"
        description: "Reference to related entity (e.g., 'Washington' for US government)"
        guideline: "Create separate entity; mark metonymic relationship"
        
      - case: "Generic Groups"
        description: "Plural generics (Musicians are creative. They practice daily.)"
        guideline: "Mark as generic; optionally exclude from scoring"

  # ---------------------------------------------------------------------------
  # EVALUATION METRICS
  # ---------------------------------------------------------------------------
  
  evaluation:
    description: |
      Metrics for evaluating coreference resolution quality.
      Multiple metrics capture different aspects of performance.      
    
    metrics:
      - name: "MUC"
        description: "Link-based F1 (Vilain et al., 1995)"
        formula: "Precision/Recall over minimum spanning tree links"
        strengths: "Penalizes missing links"
        weaknesses: "Doesn't penalize over-merging well"
        
      - name: "B-CUBED"
        description: "Mention-based precision/recall (Bagga & Baldwin, 1998)"
        formula: "Average P/R per mention based on cluster overlap"
        strengths: "Balanced for split/merge errors"
        weaknesses: "Weights all mentions equally"
        
      - name: "CEAF"
        description: "Entity-based alignment (Luo, 2005)"
        formula: "Optimal 1-1 alignment between gold and system clusters"
        variants:
          - "CEAF-e: entity-based (cluster-level)"
          - "CEAF-m: mention-based"
        strengths: "Directly evaluates entity-level accuracy"
        
      - name: "LEA"
        description: "Link-based entity-aware (Moosavi & Strube, 2016)"
        formula: "Importance-weighted link evaluation"
        strengths: "Balances entity and link perspectives"
        
      - name: "CoNLL Average"
        description: "Average of MUC, B-CUBED, CEAF-e"
        standard: "Official CoNLL shared task metric"
        formula: "(MUC_F1 + B3_F1 + CEAF_e_F1) / 3"
    
    error_types:
      - type: "False Link (Precision Error)"
        description: "System links two mentions that shouldn't be linked"
        impact: "Over-merged clusters"
        
      - type: "Missing Link (Recall Error)"
        description: "System fails to link two coreferent mentions"
        impact: "Fragmented clusters"
        
      - type: "Wrong Link"
        description: "Mention linked to wrong antecedent (but still coreferent)"
        impact: "May not affect cluster-level metrics"

  # ---------------------------------------------------------------------------
  # EXAMPLES
  # ---------------------------------------------------------------------------
  
  examples:
    - name: "Pronominal coreference"
      text: |
        Rembrandt van Rijn was born in Leiden in 1606. He moved to Amsterdam
        in 1631, where he established his studio. The Dutch master painted
        over 300 works during his career. He died in 1669.        
      
      mentions:
        - id: "m1"
          span: {start: 0, end: 18, text: "Rembrandt van Rijn"}
          type: "PROPER"
          entity_type: "AGT.PER"
        - id: "m2"
          span: {start: 48, end: 50, text: "He"}
          type: "PRONOMINAL"
          entity_type: "AGT.PER"
        - id: "m3"
          span: {start: 97, end: 100, text: "his"}
          type: "PRONOMINAL"
          entity_type: "AGT.PER"
        - id: "m4"
          span: {start: 109, end: 125, text: "The Dutch master"}
          type: "NOMINAL"
          entity_type: "AGT.PER"
        - id: "m5"
          span: {start: 157, end: 160, text: "his"}
          type: "PRONOMINAL"
          entity_type: "AGT.PER"
        - id: "m6"
          span: {start: 170, end: 172, text: "He"}
          type: "PRONOMINAL"
          entity_type: "AGT.PER"
      
      clusters:
        - cluster_id: "e1"
          entity_type: "AGT.PER"
          canonical_mention: "m1"
          mentions: ["m1", "m2", "m3", "m4", "m5", "m6"]
          external_uri: "http://viaf.org/viaf/64013650"
      
      links:
        - {source: "m2", target: "m1", type: "COREF.PRO.PER"}
        - {source: "m3", target: "m2", type: "COREF.PRO.POS"}
        - {source: "m4", target: "m1", type: "COREF.NOM.EPI"}
        - {source: "m5", target: "m4", type: "COREF.PRO.POS"}
        - {source: "m6", target: "m4", type: "COREF.PRO.PER"}

    - name: "Bridging reference"
      text: |
        The Rijksmuseum reopened in 2013 after a decade of renovation.
        The main building houses the collection of Dutch masters.
        The entrance on Museumplein welcomes 2.5 million visitors annually.        
      
      mentions:
        - id: "m1"
          span: {text: "The Rijksmuseum"}
          entity_type: "GRP.ORG"
        - id: "m2"
          span: {text: "The main building"}
          entity_type: "TOP.BLT"
        - id: "m3"
          span: {text: "The entrance"}
          entity_type: "TOP.BLT"
      
      annotations:
        - type: "COREF.BRD.PRT"
          source: "m2"
          anchor: "m1"
          relation: "part-of"
          notes: "Building is part of museum complex"
        - type: "COREF.BRD.PRT"
          source: "m3"
          anchor: "m2"
          relation: "part-of"
          notes: "Entrance is part of building"

    - name: "Split antecedent"
      text: |
        Frans Hals and Judith Leyster both painted in Haarlem.
        They influenced each other's work.        
      
      mentions:
        - id: "m1"
          span: {text: "Frans Hals"}
          entity_type: "AGT.PER"
        - id: "m2"
          span: {text: "Judith Leyster"}
          entity_type: "AGT.PER"
        - id: "m3"
          span: {text: "They"}
          entity_type: "AGT.GRP"
        - id: "m4"
          span: {text: "each other's"}
          entity_type: "AGT.PER"
      
      annotations:
        - type: "COREF.SPL"
          source: "m3"
          antecedents: ["m1", "m2"]
          notes: "Plural pronoun with split antecedent"
        - type: "COREF.PRO.REC"
          source: "m4"
          target: "m3"
          notes: "Reciprocal within split-antecedent group"

# =============================================================================
# SECTION 18: UNCERTAINTY AND CONFIDENCE SCORING
# =============================================================================
# 
# This section defines the framework for expressing and scoring uncertainty
# in entity annotations, relationships, and coreference decisions.
#
# Uncertainty is inherent in NLP/NER and must be explicitly modeled to:
#   - Prioritize human review (low confidence → needs verification)
#   - Enable probabilistic reasoning in downstream applications
#   - Support provenance tracking (who asserted what, with what certainty)
#   - Allow confidence-based filtering for different use cases
#
# Ontological foundations:
#   - W3C PROV-O: Provenance tracking
#   - OA (Open Annotation): Annotation confidence
#   - CIDOC-CRM: E13 Attribute Assignment for epistemic claims
#   - IAO (Information Artifact Ontology): Information quality
#   - ISO 19157: Data quality for geographic information
#   - TEI: @cert (certainty) and <certainty> element
#   - TimeML: @temporalFunction for temporal uncertainty
#
# Statistical foundations:
#   - Bayesian probability for belief updating
#   - Calibration theory for reliable confidence scores
#   - Inter-annotator agreement metrics (Cohen's κ, Fleiss' κ)
# =============================================================================

uncertainty_and_confidence:
  
  description: |
    Uncertainty modeling captures the degree of confidence in annotations,
    enabling consumers to filter, prioritize, and reason about data quality.
    
    This framework addresses THREE types of uncertainty:
    
    1. ANNOTATION UNCERTAINTY: Confidence in a specific annotation decision
       - "How sure are we that 'Rembrandt' refers to AGT.PER?"
       
    2. EPISTEMIC UNCERTAINTY: Uncertainty about facts in the world
       - "Did Rembrandt actually paint this work?" (attribution uncertainty)
       
    3. LINGUISTIC UNCERTAINTY: Ambiguity in the source text itself
       - "The text says 'possibly by Rembrandt'" (hedged language)
    
    These are distinct and must be tracked separately.    
  
  namespaces:
    prov: "http://www.w3.org/ns/prov#"
    oa: "http://www.w3.org/ns/oa#"
    crm: "http://www.cidoc-crm.org/cidoc-crm/"
    iao: "http://purl.obolibrary.org/obo/IAO_"
    dqv: "http://www.w3.org/ns/dqv#"
    tei: "http://www.tei-c.org/ns/1.0"
    schema: "http://schema.org/"
    xsd: "http://www.w3.org/2001/XMLSchema#"

  # ---------------------------------------------------------------------------
  # CONFIDENCE SCORE FRAMEWORK
  # ---------------------------------------------------------------------------
  
  confidence_framework:
    description: |
      Numerical confidence scores on [0.0, 1.0] scale with defined semantics.
      Scores should be CALIBRATED: a score of 0.8 means the annotation is
      correct approximately 80% of the time across similar cases.      
    
    score_semantics:
      - range: "[0.95, 1.0]"
        range_min: 0.95
        range_max: 1.0
        inclusive: "both"
        label: "CERTAIN"
        tei_cert: "high"
        description: |
          Near-certain annotation. Human annotators would agree 95%+ of the time.
          Examples: exact string matches, unambiguous proper nouns, rule-based
          high-precision extractions.          
        action: "Accept automatically; no review needed"
        color_code: "#228B22"  # Forest green
        
      - range: "[0.80, 0.95)"
        range_min: 0.80
        range_max: 0.95
        inclusive: "left"
        label: "HIGH_CONFIDENCE"
        tei_cert: "high"
        description: |
          High confidence annotation. Strong signals but some ambiguity possible.
          Examples: well-known entities with standard forms, clear context.          
        action: "Accept with spot-checking"
        color_code: "#32CD32"  # Lime green
        
      - range: "[0.60, 0.80)"
        range_min: 0.60
        range_max: 0.80
        inclusive: "left"
        label: "MEDIUM_CONFIDENCE"
        tei_cert: "medium"
        description: |
          Moderate confidence. Multiple interpretations possible; context helps.
          Examples: common names with multiple referents, unclear boundaries.          
        action: "Review if resource-sensitive; accept for bulk processing"
        color_code: "#FFD700"  # Gold
        
      - range: "[0.40, 0.60)"
        range_min: 0.40
        range_max: 0.60
        inclusive: "left"
        label: "LOW_CONFIDENCE"
        tei_cert: "low"
        description: |
          Low confidence. Significant uncertainty; near chance for binary decision.
          Examples: ambiguous pronouns, unknown entities, noisy text.          
        action: "Flag for human review"
        color_code: "#FFA500"  # Orange
        
      - range: "[0.20, 0.40)"
        range_min: 0.20
        range_max: 0.40
        inclusive: "left"
        label: "VERY_LOW_CONFIDENCE"
        tei_cert: "low"
        description: |
          Very low confidence. Model is guessing; alternative interpretations likely.          
        action: "Require human verification before use"
        color_code: "#FF4500"  # Orange-red
        
      - range: "[0.0, 0.20)"
        range_min: 0.0
        range_max: 0.20
        inclusive: "left"
        label: "UNCERTAIN"
        tei_cert: "unknown"
        description: |
          Near-zero confidence. Essentially no signal; included for completeness.          
        action: "Do not use without manual annotation"
        color_code: "#DC143C"  # Crimson
    
    special_values:
      - value: null
        meaning: "Confidence not computed/available"
        
      - value: 1.0
        meaning: "Absolute certainty (use sparingly; only for definitional truths)"
        
      - value: 0.0
        meaning: "Definite negative (annotation is known to be incorrect)"
    
    aggregation_rules:
      description: "How to combine confidence scores"
      
      rules:
        - name: "Independent Conjunction"
          formula: "P(A ∧ B) = P(A) × P(B)"
          use_case: "Joint probability of independent annotations"
          example: "Entity type AND entity boundaries both correct"
          
        - name: "Conservative Minimum"
          formula: "min(conf_1, conf_2, ...)"
          use_case: "Weakest-link scenarios; overall quality limited by weakest component"
          example: "Relationship confidence = min(subject_conf, object_conf, predicate_conf)"
          
        - name: "Weighted Average"
          formula: "Σ(weight_i × conf_i) / Σ(weight_i)"
          use_case: "Combining scores from multiple sources/annotators"
          example: "Ensemble model output"
          
        - name: "Maximum"
          formula: "max(conf_1, conf_2, ...)"
          use_case: "Any-correct scenarios (at least one interpretation valid)"
          example: "Multiple valid entity types"

  # ---------------------------------------------------------------------------
  # ANNOTATION UNCERTAINTY
  # ---------------------------------------------------------------------------
  
  annotation_uncertainty:
    description: |
      Uncertainty about the correctness of annotation DECISIONS.
      This is process uncertainty, not world uncertainty.      
    
    dimensions:
      - dimension: "boundary_confidence"
        description: "Confidence in span boundaries (start/end offsets)"
        factors:
          - "Tokenization ambiguity"
          - "Nested entity decisions"
          - "Modifier attachment"
        examples:
          - text: "the Dutch painter Rembrandt"
            issue: "Should span include 'the Dutch painter' or just 'Rembrandt'?"
            resolution: "Mark 'Rembrandt' as head; broader span has lower confidence"
            
      - dimension: "type_confidence"
        description: "Confidence in entity type assignment"
        factors:
          - "Ambiguous entity types (organization vs. event: 'the conference')"
          - "Metonymy (place for organization: 'Washington announced')"
          - "Type granularity (museum subtype assignment)"
        examples:
          - text: "Apple announced new products"
            issue: "AGT.PER (person)? GRP.COR (corporation)? WRK.OBJ (fruit)?"
            resolution: "Context disambiguates; assign GRP.COR with high confidence"
            
      - dimension: "referent_confidence"
        description: "Confidence in identity of referred entity"
        factors:
          - "Ambiguous names (multiple people named 'John Smith')"
          - "Unknown entities (not in knowledge base)"
          - "Nickname/variant resolution"
        examples:
          - text: "Dr. Williams presented the findings"
            issue: "Which Dr. Williams? No disambiguating context."
            resolution: "Create entity with low referent_confidence; flag for KB linking"
            
      - dimension: "extraction_confidence"
        description: "Confidence from the extraction model/process"
        factors:
          - "Model probability/logits"
          - "Rule match specificity"
          - "OCR quality (for digitized documents)"
    
    combined_score:
      formula: |
        annotation_confidence = min(
          boundary_confidence,
          type_confidence,
          referent_confidence
        ) × extraction_confidence        
      
      rationale: |
        Use minimum for boundary/type/referent (weakest link), then scale by
        extraction quality. Low OCR confidence degrades all downstream scores.        

  # ---------------------------------------------------------------------------
  # EPISTEMIC UNCERTAINTY
  # ---------------------------------------------------------------------------
  
  epistemic_uncertainty:
    description: |
      Uncertainty about FACTS IN THE WORLD, not annotation process.
      Captures disputed, unknown, or probability claims about reality.      
    
    uncertainty_types:
      - type: "DISPUTED"
        description: "Multiple conflicting claims in sources"
        examples:
          - "Scholars dispute whether Vermeer used a camera obscura"
          - "The painting's attribution is contested"
        annotation_fields:
          - field: "epistemic_status"
            value: "disputed"
          - field: "competing_claims"
            value: ["claim_1", "claim_2"]
           
      - type: "UNKNOWN"
        description: "No reliable information available"
        examples:
          - "The artist's birthdate is unknown"
          - "The work's provenance before 1900 is unrecorded"
        annotation_fields:
          - field: "epistemic_status"
            value: "unknown"
           
      - type: "APPROXIMATE"
        description: "Value is estimated or rounded"
        examples:
          - "The collection contains approximately 8,000 works"
          - "Painted circa 1642"
        annotation_fields:
          - field: "epistemic_status"
            value: "approximate"
          - field: "precision"
            value: "circa_year | circa_decade | order_of_magnitude"
           
      - type: "INFERRED"
        description: "Derived from other facts, not directly stated"
        examples:
          - "If born in 1606 and died in 1669, he lived 63 years"
          - "Based on style, attributed to the artist's late period"
        annotation_fields:
          - field: "epistemic_status"
            value: "inferred"
          - field: "inference_basis"
            value: "description of reasoning"
           
      - type: "HYPOTHETICAL"
        description: "Conditional or speculative claim"
        examples:
          - "If the signature is authentic, the painting is by Vermeer"
          - "The proposed identification remains unconfirmed"
        annotation_fields:
          - field: "epistemic_status"
            value: "hypothetical"
          - field: "condition"
            value: "description of condition"
    
    source_reliability:
      description: "Confidence based on source quality"
      
      tiers:
        - tier: 1
          label: "AUTHORITATIVE"
          sources:
            - "Primary sources (original documents, eyewitness accounts)"
            - "Official registries (ISIL, Wikidata with references)"
            - "Peer-reviewed scholarship"
          default_confidence: 0.95
          
        - tier: 2
          label: "RELIABLE"
          sources:
            - "Institutional websites"
            - "Encyclopedia entries (Britannica, Wikipedia with citations)"
            - "Expert secondary sources"
          default_confidence: 0.85
          
        - tier: 3
          label: "CREDIBLE"
          sources:
            - "News media"
            - "Wikipedia without citations"
            - "Aggregated databases"
          default_confidence: 0.70
          
        - tier: 4
          label: "UNVERIFIED"
          sources:
            - "User-generated content"
            - "Social media"
            - "NLP extraction without verification"
          default_confidence: 0.50
          
        - tier: 5
          label: "SUSPECT"
          sources:
            - "Known unreliable sources"
            - "Contradicted by authoritative sources"
            - "Outdated information"
          default_confidence: 0.20

  # ---------------------------------------------------------------------------
  # LINGUISTIC UNCERTAINTY
  # ---------------------------------------------------------------------------
  
  linguistic_uncertainty:
    description: |
      Uncertainty encoded IN THE SOURCE TEXT itself through hedging,
      modality, attribution, and evidentiality markers.      
    
    hedging_markers:
      description: "Words/phrases indicating reduced certainty in source"
      
      categories:
        - category: "MODAL_VERBS"
          markers:
            - marker: "may/might"
              uncertainty_reduction: 0.4
              example: "The painting may be by Vermeer"
            - marker: "could"
              uncertainty_reduction: 0.3
              example: "This could date to the 1640s"
            - marker: "must (epistemic)"
              uncertainty_reduction: 0.1
              example: "He must have known the artist"
            - marker: "should"
              uncertainty_reduction: 0.2
              example: "The document should contain the date"
              
        - category: "HEDGING_ADVERBS"
          markers:
            - marker: "possibly/perhaps/maybe"
              uncertainty_reduction: 0.4
              example: "Possibly painted in Amsterdam"
            - marker: "probably/likely"
              uncertainty_reduction: 0.2
              example: "Probably by Rembrandt's workshop"
            - marker: "certainly/definitely"
              uncertainty_reduction: 0.0
              example: "Certainly authentic"
            - marker: "apparently/seemingly"
              uncertainty_reduction: 0.3
              example: "Apparently a self-portrait"
              
        - category: "HEDGING_ADJECTIVES"
          markers:
            - marker: "possible/potential"
              uncertainty_reduction: 0.4
              example: "A possible attribution to Hals"
            - marker: "probable/likely"
              uncertainty_reduction: 0.2
              example: "The probable author"
            - marker: "alleged/purported"
              uncertainty_reduction: 0.5
              example: "The alleged forgery"
            - marker: "so-called"
              uncertainty_reduction: 0.3
              example: "The so-called 'Night Watch'"
              
        - category: "ATTRIBUTION_PHRASES"
          description: "Attribution to uncertain source"
          markers:
            - marker: "according to X"
              uncertainty_reduction: "depends on X reliability"
              example: "According to early sources..."
            - marker: "it is said that"
              uncertainty_reduction: 0.4
              example: "It is said that Vermeer used..."
            - marker: "traditionally attributed to"
              uncertainty_reduction: 0.3
              example: "Traditionally attributed to Leonardo"
            - marker: "some scholars believe"
              uncertainty_reduction: 0.3
              example: "Some scholars believe this is..."
              
        - category: "APPROXIMATION"
          markers:
            - marker: "circa/c./ca."
              uncertainty_reduction: 0.2
              example: "c. 1642"
            - marker: "approximately/about/around"
              uncertainty_reduction: 0.2
              example: "Around 1,000 works"
            - marker: "roughly/nearly"
              uncertainty_reduction: 0.2
              example: "Nearly 50 years old"
    
    annotation_pattern:
      description: "How to annotate linguistic uncertainty"
      
      schema:
        - field: "has_hedging"
          type: "boolean"
          description: "True if source text contains hedging"
          
        - field: "hedging_markers"
          type: "array[string]"
          description: "List of hedging markers detected"
          
        - field: "source_certainty"
          type: "float"
          description: "Certainty expressed in source (before extraction adjustment)"
          formula: "1.0 - max(uncertainty_reductions)"
          
        - field: "attributed_to"
          type: "object"
          description: "If claim is attributed, to whom"
          schema:
            source_name: "string"
            source_reliability: "float"
            is_author_endorsement: "boolean"
      
      examples:
        - text: "The painting is possibly by Rembrandt"
          annotation:
            has_hedging: true
            hedging_markers: ["possibly"]
            source_certainty: 0.6
            notes: "Attribution uncertain in source text"
            
        - text: "According to Houbraken, Rembrandt was born in Leiden"
          annotation:
            has_hedging: true
            hedging_markers: ["according to"]
            attributed_to:
              source_name: "Arnold Houbraken"
              source_reliability: 0.8
              is_author_endorsement: false
            source_certainty: 0.8

  # ---------------------------------------------------------------------------
  # UNCERTAINTY ANNOTATION SCHEMA
  # ---------------------------------------------------------------------------
  
  annotation_schema:
    description: "Complete schema for uncertainty annotation"
    
    fields:
      # Core confidence score
      - field: "confidence"
        type: "float"
        range: [0.0, 1.0]
        required: true
        description: "Overall confidence in the annotation"
        
      # Detailed confidence breakdown
      - field: "confidence_breakdown"
        type: "object"
        optional: true
        schema:
          boundary_confidence: "float"
          type_confidence: "float"
          referent_confidence: "float"
          extraction_confidence: "float"
        description: "Component confidence scores"
        
      # Confidence metadata
      - field: "confidence_method"
        type: "enum"
        values:
          - "MODEL_PROBABILITY"    # From ML model output
          - "RULE_CONFIDENCE"      # Assigned by rule
          - "HUMAN_JUDGMENT"       # Manual annotation
          - "ENSEMBLE_AGGREGATE"   # Combined from multiple sources
          - "HEURISTIC"           # Based on heuristic rules
          - "CALIBRATED"          # Post-hoc calibrated
        description: "How confidence was computed"
        
      - field: "calibration_status"
        type: "enum"
        values:
          - "UNCALIBRATED"    # Raw model output
          - "CALIBRATED"      # Adjusted for reliability
          - "VALIDATED"       # Verified against ground truth
        description: "Whether score has been calibrated"
        
      # Epistemic uncertainty
      - field: "epistemic_status"
        type: "enum"
        values:
          - "ASSERTED"        # Presented as fact
          - "DISPUTED"        # Multiple conflicting claims
          - "UNKNOWN"         # No reliable information
          - "APPROXIMATE"     # Estimated value
          - "INFERRED"        # Derived from other facts
          - "HYPOTHETICAL"    # Conditional/speculative
        optional: true
        default: "ASSERTED"
        description: "Epistemic status of the claim"
        
      - field: "competing_claims"
        type: "array[object]"
        optional: true
        description: "If disputed, list of competing claims"
        item_schema:
          claim: "string"
          source: "string"
          confidence: "float"
        
      # Linguistic uncertainty
      - field: "source_hedging"
        type: "object"
        optional: true
        schema:
          has_hedging: "boolean"
          markers: "array[string]"
          source_certainty: "float"
        description: "Hedging detected in source text"
        
      # Attribution
      - field: "attribution"
        type: "object"
        optional: true
        schema:
          attributed_to: "string (source/speaker)"
          attribution_type: "enum (QUOTE, PARAPHRASE, CLAIM)"
          endorsement: "boolean (does author endorse?)"
        description: "If claim is attributed to another source"
        
      # Review status
      - field: "review_status"
        type: "enum"
        values:
          - "UNREVIEWED"      # Not yet reviewed
          - "PENDING_REVIEW"  # Flagged for review
          - "REVIEWED"        # Reviewed, accepted
          - "DISPUTED"        # Reviewer disagrees
          - "CORRECTED"       # Original was wrong, corrected
        optional: true
        description: "Human review status"
        
      - field: "reviewed_by"
        type: "string"
        optional: true
        description: "Reviewer identifier"
        
      - field: "review_date"
        type: "string"
        format: "ISO 8601"
        optional: true
        description: "When reviewed"
        
      - field: "review_notes"
        type: "string"
        optional: true
        description: "Reviewer comments"

  # ---------------------------------------------------------------------------
  # CALIBRATION AND VALIDATION
  # ---------------------------------------------------------------------------
  
  calibration:
    description: |
      Methods for ensuring confidence scores are RELIABLE (calibrated).
      A calibrated score of 0.8 means 80% of annotations at that score are correct.      
    
    calibration_methods:
      - method: "Temperature Scaling"
        description: |
          Post-hoc calibration using held-out validation set.
          Learn temperature T such that softmax(logits/T) is calibrated.          
        suitable_for: "Neural model outputs"
        
      - method: "Platt Scaling"
        description: |
          Fit sigmoid function to map raw scores to calibrated probabilities.
          P(correct) = 1 / (1 + exp(A × score + B))          
        suitable_for: "Binary classification outputs"
        
      - method: "Isotonic Regression"
        description: |
          Non-parametric calibration preserving score ordering.
          Maps scores to calibrated values via piecewise constant function.          
        suitable_for: "When calibration curve is non-monotonic"
        
      - method: "Histogram Binning"
        description: |
          Bin predictions and assign calibrated score per bin.          
        suitable_for: "Simple, interpretable calibration"
        
    calibration_metrics:
      - metric: "Expected Calibration Error (ECE)"
        description: "Average gap between confidence and accuracy per bin"
        formula: "Σ |B_i|/n × |accuracy(B_i) - confidence(B_i)|"
        target: "< 0.05"
        
      - metric: "Maximum Calibration Error (MCE)"
        description: "Largest gap across bins"
        formula: "max_i |accuracy(B_i) - confidence(B_i)|"
        target: "< 0.10"
        
      - metric: "Brier Score"
        description: "Mean squared error of probability estimates"
        formula: "Σ (p_i - y_i)² / n"
        target: "Lower is better"
    
    validation_requirements:
      description: "Requirements for validating confidence scores"
      
      requirements:
        - "Held-out test set with ground truth annotations"
        - "Stratified sampling across entity types and confidence ranges"
        - "Minimum 100 samples per calibration bin"
        - "Regular recalibration as model/data changes"
      
      monitoring:
        - "Track calibration metrics over time"
        - "Alert on calibration drift"
        - "Retrain calibration when ECE exceeds threshold"

  # ---------------------------------------------------------------------------
  # INTER-ANNOTATOR AGREEMENT
  # ---------------------------------------------------------------------------
  
  inter_annotator_agreement:
    description: |
      Metrics for measuring agreement between annotators, which informs
      confidence estimation and task difficulty assessment.      
    
    metrics:
      - metric: "Cohen's Kappa (κ)"
        description: "Agreement corrected for chance (2 annotators)"
        formula: "κ = (P_o - P_e) / (1 - P_e)"
        interpretation:
          - range: [0.81, 1.0]
            label: "Almost perfect"
          - range: [0.61, 0.80]
            label: "Substantial"
          - range: [0.41, 0.60]
            label: "Moderate"
          - range: [0.21, 0.40]
            label: "Fair"
          - range: [0.0, 0.20]
            label: "Slight"
        
      - metric: "Fleiss' Kappa"
        description: "Multi-annotator extension of Cohen's κ"
        use_case: "3+ annotators on same items"
        
      - metric: "Krippendorff's Alpha (α)"
        description: "Handles missing data, any number of annotators"
        use_case: "Production annotation with variable annotator coverage"
        
      - metric: "F1 Agreement"
        description: "Treat one annotator as gold, compute F1"
        use_case: "When one annotator is more senior/authoritative"
    
    agreement_to_confidence:
      description: |
        Use IAA to inform confidence scoring. Low agreement items
        should have lower confidence bounds.        
      
      heuristic:
        - agreement: "All annotators agree"
          confidence_boost: 0.1
          
        - agreement: "Majority agrees (>66%)"
          confidence_boost: 0.0
          
        - agreement: "Split decision (50%)"
          confidence_reduction: 0.2
          flag: "REQUIRES_ADJUDICATION"
          
        - agreement: "No majority"
          confidence_reduction: 0.4
          flag: "HIGHLY_AMBIGUOUS"

  # ---------------------------------------------------------------------------
  # UNCERTAINTY PROPAGATION
  # ---------------------------------------------------------------------------
  
  uncertainty_propagation:
    description: |
      How uncertainty flows through annotation pipelines and affects
      downstream tasks.      
    
    propagation_rules:
      - stage: "Entity Extraction → Relationship Extraction"
        rule: |
          Relationship confidence ≤ min(subject_confidence, object_confidence)          
        rationale: "Can't have high-confidence relationship with low-confidence entities"
        
      - stage: "Mention Detection → Coreference"
        rule: |
          Coreference confidence ≤ min(mention_1_confidence, mention_2_confidence)          
        rationale: "Coreference uncertain if mentions are uncertain"
        
      - stage: "Individual Annotations → Aggregate Statistics"
        rule: |
          Report confidence intervals, not just point estimates.
          E.g., "8,000 ± 500 entities extracted (95% CI)"          
        
      - stage: "Multiple Sources → Merged Entity"
        rule: |
          merged_confidence = f(source_confidences, agreement)
          Agreement boosts; conflict reduces.          
    
    monte_carlo_simulation:
      description: |
        For complex pipelines, use Monte Carlo simulation:
        1. Sample annotations according to confidence distributions
        2. Run downstream pipeline
        3. Aggregate to get output distribution        
      use_case: "Estimating uncertainty in knowledge graph population"

  # ---------------------------------------------------------------------------
  # PRESENTATION AND VISUALIZATION
  # ---------------------------------------------------------------------------
  
  presentation:
    description: "How to present uncertainty to users"
    
    ui_guidelines:
      - guideline: "Color Coding"
        description: "Use consistent color scale from green (certain) to red (uncertain)"
        implementation:
          - confidence_range: "[0.9, 1.0]"
            confidence_min: 0.9
            confidence_max: 1.0
            color: "green"
            icon: "✓"
          - confidence_range: "[0.7, 0.9)"
            confidence_min: 0.7
            confidence_max: 0.9
            color: "light-green"
            icon: "○"
          - confidence_range: "[0.5, 0.7)"
            confidence_min: 0.5
            confidence_max: 0.7
            color: "yellow"
            icon: "?"
          - confidence_range: "[0.3, 0.5)"
            confidence_min: 0.3
            confidence_max: 0.5
            color: "orange"
            icon: "⚠"
          - confidence_range: "[0.0, 0.3)"
            confidence_min: 0.0
            confidence_max: 0.3
            color: "red"
            icon: "✗"
            
      - guideline: "Uncertainty Indicators"
        description: "Visual indicators scaled to confidence"
        options:
          - "Border thickness (thicker = less certain)"
          - "Opacity (more transparent = less certain)"
          - "Hatching/patterns for uncertain regions"
          - "Tooltip with detailed confidence breakdown"
          
      - guideline: "Sortable/Filterable"
        description: "Allow users to sort and filter by confidence"
        features:
          - "Slider to set confidence threshold"
          - "Show only items needing review (conf < 0.8)"
          - "Sort by confidence ascending (most uncertain first)"
          
      - guideline: "Confidence Distribution"
        description: "Show overall confidence distribution"
        implementation: "Histogram of confidence scores across annotations"
        
      - guideline: "Explain Uncertainty"
        description: "On hover/click, explain WHY confidence is low"
        example: |
          "Low confidence (0.45) because:
           - Ambiguous entity type (0.52)
           - Multiple possible referents in KB (0.65)
           - Source text contains hedging: 'possibly'"          

  # ---------------------------------------------------------------------------
  # EXAMPLES
  # ---------------------------------------------------------------------------
  
  examples:
    - name: "High confidence entity"
      text: "The Rijksmuseum in Amsterdam houses Rembrandt's Night Watch."
      annotation:
        entity: "Rijksmuseum"
        entity_type: "GRP.ORG"
        confidence: 0.97
        confidence_breakdown:
          boundary_confidence: 0.99
          type_confidence: 0.98
          referent_confidence: 0.96
          extraction_confidence: 0.98
        confidence_method: "MODEL_PROBABILITY"
        epistemic_status: "ASSERTED"
        external_uri: "https://www.wikidata.org/entity/Q190804"
        
    - name: "Hedged attribution"
      text: "The painting is possibly by Rembrandt or his workshop."
      annotation:
        entity: "Rembrandt or his workshop"
        entity_type: "AGT.PER"
        confidence: 0.75
        confidence_breakdown:
          boundary_confidence: 0.85
          type_confidence: 0.90
          referent_confidence: 0.60
          extraction_confidence: 0.85
        source_hedging:
          has_hedging: true
          markers: ["possibly", "or"]
          source_certainty: 0.5
        epistemic_status: "DISPUTED"
        competing_claims:
          - claim: "By Rembrandt himself"
            confidence: 0.4
          - claim: "By Rembrandt's workshop"
            confidence: 0.5
          - claim: "Later copy"
            confidence: 0.1
            
    - name: "Approximate temporal reference"
      text: "The collection was established around 1800."
      annotation:
        entity: "around 1800"
        entity_type: "TMP.DAB"
        confidence: 0.85
        epistemic_status: "APPROXIMATE"
        source_hedging:
          has_hedging: true
          markers: ["around"]
          source_certainty: 0.8
        temporal_uncertainty:
          point_estimate: "1800-01-01"
          range_start: "1795-01-01"
          range_end: "1805-12-31"
          precision: "circa_decade"
          
    - name: "Low confidence extraction needing review"
      text: "Dr. J. van der Berg described the artifact."
      annotation:
        entity: "Dr. J. van der Berg"
        entity_type: "AGT.PER"
        confidence: 0.45
        confidence_breakdown:
          boundary_confidence: 0.90
          type_confidence: 0.85
          referent_confidence: 0.35
          extraction_confidence: 0.80
        confidence_method: "MODEL_PROBABILITY"
        review_status: "PENDING_REVIEW"
        review_notes: |
          Multiple "J. van der Berg" in knowledge base.
          Need additional context for disambiguation.          
        possible_referents:
          - uri: "https://viaf.org/viaf/12345"
            name: "Johan van der Berg (1890-1960)"
            match_confidence: 0.35
          - uri: "https://viaf.org/viaf/67890"
            name: "Johannes van der Berg (1920-1995)"
            match_confidence: 0.30
          - uri: null
            name: "(Unknown person)"
            match_confidence: 0.35

# =============================================================================
# END OF CONVENTION
# =============================================================================
# 
# This convention is COMPLETE and SELF-CONTAINED.
# Version: 1.7.0-unified
# Date: 2025-12-02
# 
# For questions or updates, see: docs/convention/README.md
# =============================================================================