kempersc dfb4744dc7 Evaluate data enrichments of persons

2026-01-11 12:15:27 +01:00

153 KiB

Raw Blame History

AI Agent Instructions for GLAM Data Extraction

This document provides instructions for AI agents (particularly OpenCODE and Claude) to assist with extracting heritage institution data from conversation JSON files and other sources.

🎯 PROJECT CORE MISSION

PRIMARY OBJECTIVE: Create a comprehensive, nuanced ontology that accurately represents the complex, temporal, multi-faceted nature of heritage custodian institutions worldwide.

This is NOT a simple data extraction project. This is an ontology engineering project that:

Models heritage entities as multi-aspect temporal entities (place, custodian, legal form, collections, people)
Integrates multiple base ontologies (CPOV, TOOI, CIDOC-CRM, RiC-O, Schema.org, PiCo)
Captures organizational change events over time (custody transfers, mergers, transformations)
Distinguishes between nominal references and formal organizational structures
Links heritage custodians to people, collections, and locations with independent temporal lifecycles

If you're looking for simple NER extraction, this is not the right project.

🚨 CRITICAL RULES FOR ALL AGENTS

⚠️ DATA QUALITY IS OF UTMOST IMPORTANCE ⚠️

Wrong data is worse than no data. All enrichments MUST be double-checked before being committed to the dataset. A single false claim (birth year from wrong person, social media from random account) corrupts the entire profile.

Mandatory verification for ALL enrichments:

Source verification: Is the source about the SAME person, not a namesake?
Cross-reference check: Do at least 3 identity attributes match (employer, location, profession, age, education)?
Conflict detection: Any conflicting signal (actress vs curator, Venezuela vs UK) = REJECT
Provenance documentation: Every claim must have full provenance with retrieval timestamp and source URL

If in doubt, DO NOT add the claim. It is better to have incomplete data than incorrect data.

🚫 AUTOMATED WEB ENRICHMENT IS PROHIBITED 🚫

DO NOT USE automated scripts to enrich person profiles with web search data. The enrich_person_comprehensive.py script has been DEPRECATED due to catastrophic entity resolution failures.

What happened: Automated enrichment attributed data from wrong people with similar names:

Birth year from Venezuelan actress attributed to UK art curator
ResearchGate profile of Mexican hydrogeologist attributed to Dutch museum curator
Wikipedia article about Nazi doctor attributed to heritage worker
540+ false claims had to be manually removed

ALL person enrichment must be done MANUALLY with human verification that the source refers to the correct person.

This section summarizes 47 critical rules. Each rule has complete documentation in .opencode/ files.

Rule 0: LinkML Schemas Are the Single Source of Truth

🚨 CRITICAL: LinkML schema files in schemas/20251121/linkml/ are the authoritative definition of the Heritage Custodian Ontology.

Key Points:

ALL derived files (RDF, TypeDB, UML) are GENERATED - never edit them directly
Always use full timestamps (YYYYMMDD_HHMMSS) in generated filenames
Primary schema: schemas/20251121/linkml/01_custodian_name.yaml

Workflow:

1. EDIT LinkML schema
2. REGENERATE: gen-owl → rdfpipe → all 8 RDF formats
3. REGENERATE: gen-yuml → UML diagrams
4. UPDATE: TypeDB schema (manual)
5. VALIDATE: linkml-validate

See: .opencode/SCHEMA_GENERATION_RULES.md for complete generation rules

Rule 0b: LinkML Type/Types File Naming Convention

🚨 CRITICAL: When creating class hierarchies that replace enums, follow the Type/Types naming pattern.

Naming Pattern:

[Entity]Type.yaml (singular): Abstract base class defining the type taxonomy
[Entity]Types.yaml (plural): File containing all concrete subclasses

Examples:

Base Class File	Subclasses File	Description
`DigitalPlatformType.yaml`	`DigitalPlatformTypes.yaml`	Digital platform type taxonomy (69 types)
`WebPortalType.yaml`	`WebPortalTypes.yaml`	Web portal type taxonomy
`CustodianType.yaml`	`CustodianTypes.yaml`	Heritage custodian type taxonomy (GLAMORCUBESFIXPHDNT)

Import Pattern:

# In DigitalPlatformTypes.yaml (subclasses file)
imports:
  - ./DigitalPlatformType  # Import base class

classes:
  DigitalLibrary:
    is_a: DigitalPlatformType  # Inherit from base
    # ...

Rationale:

Clarity: "Type" (singular) = one abstract concept; "Types" (plural) = many concrete subclasses
Discoverability: Related files are adjacent in directory listings
Consistency: Follows established pattern across schema (CustodianType/CustodianTypes, WebPortalType/WebPortalTypes)

Anti-Pattern:

❌ DigitalPlatformTypeBase.yaml - "Base" suffix is redundant; use singular "Type" instead
❌ DigitalPlatformTypeClasses.yaml - "Classes" is less intuitive than "Types"

See: .opencode/rules/type-naming-convention.md for complete documentation

Rule 1: Ontology Files Are Your Primary Reference

🚨 CRITICAL: Before designing any schema, class, or property, consult base ontologies.

Required Steps:

READ base ontology files in /data/ontology/
SEARCH for existing classes and properties
DOCUMENT your ontology alignment with rationale
NEVER invent custom properties when ontology equivalents exist

Available Ontologies:

tooiont.ttl - TOOI (Dutch government)
core-public-organisation-ap.ttl - CPOV (EU public sector)
schemaorg.owl - Schema.org (web semantics)
CIDOC_CRM_v7.1.3.rdf - CIDOC-CRM (cultural heritage)
RiC-O_1-1.rdf - Records in Contexts (archival)
pico.ttl - PiCo (person observations)

See: .opencode/HYPER_MODULAR_STRUCTURE.md for complete documentation

Rule 2: Wikidata Entities Are NOT Ontology Classes

🚨 CRITICAL: Files in data/wikidata/GLAMORCUBEPSXHFN/ contain Wikidata Q-numbers for institution TYPES, NOT formal ontology class definitions.

Workflow: Wikidata Q-number → Analyze semantics → Search ontologies → Map to ontology class → Document rationale

Note: Full rule content preserved in Appendix below (no .opencode equivalent).

Rule 3: Multi-Aspect Modeling is Mandatory

🚨 CRITICAL: Every heritage entity has MULTIPLE ontological aspects with INDEPENDENT temporal lifecycles.

Required Aspects:

Aspect	Ontology Class	Temporal Example
Place	`crm:E27_Site`	Building: 1880-present
Custodian	`cpov:PublicOrganisation`	Foundation: 1994-present
Legal Form	`org:FormalOrganization`	Registration: 1994-present
Collections	`rico:RecordSet`	Accession dates vary
People	`pico:PersonObservation`	Employment: 2020-present
Events	`crm:E10_Transfer_of_Custody`	Discrete timestamps

Note: Full rule content preserved in Appendix below (no .opencode equivalent).

Rule 4: Technical Classes Are Excluded from Visualizations

🚨 CRITICAL: Some LinkML classes exist solely for validation (e.g., Container with tree_root: true). These have NO semantic significance and MUST be excluded from UML diagrams.

Excluded Classes: Container (tree_root for validation only)

See: .opencode/LINKML_TECHNICAL_CLASSES.md for complete documentation

Rule 5: NEVER Delete Enriched Data - Additive Only

🚨 CRITICAL: Data enrichment is ADDITIVE ONLY. Never delete or overwrite existing enriched content.

Protected Data Types:

Source	Protected Fields
Google Maps	`reviews`, `rating`, `photo_count`, `popular_times`, `place_id`
OpenStreetMap	`osm_id`, `osm_type`, `osm_tags`, `amenity`, `heritage`
Wikidata	`wikidata_id`, `claims`, `sitelinks`, `aliases`
Website Scrape	`organization_details`, `collections`, `contact`, `social_media`
ISIL Registry	`isil_code`, `assigned_date`, `remarks`

See: .opencode/DATA_PRESERVATION_RULES.md for complete documentation

Rule 6: WebObservation Claims MUST Have XPath Provenance

🚨 CRITICAL: Every claim extracted from a webpage MUST have an XPath pointer to the exact location in archived HTML. Claims without XPath provenance are FABRICATED.

Required Fields:

claim_type: full_name
claim_value: "Institution Name"
source_url: https://example.org/about
retrieved_on: "2025-11-29T12:28:00Z"
xpath: /html/body/div[1]/h1
html_file: web/GHCID/example.org/rendered.html
xpath_match_score: 1.0

Scope: Applies to WebClaim and WebObservation classes. Other classes (CustodianTimelineEvent, GoogleMapsEnrichment) have different provenance models.

See: .opencode/WEB_OBSERVATION_PROVENANCE_RULES.md for complete documentation

Rule 7: Deployment is LOCAL via SSH/rsync (NO CI/CD)

🚨 CRITICAL: NO GitHub Actions. ALL deployments executed locally via SSH and rsync.

Server: 91.98.224.44 (Hetzner Cloud)

Two Frontend Apps (MONOREPO):

Domain	Local Directory	Server Path
bronhouder.nl	`/frontend/`	`/var/www/glam-frontend/`
archief.support	`/apps/archief-assistent/`	`/var/www/archief-assistent/`

Deployment Commands:

./infrastructure/deploy.sh --frontend  # bronhouder.nl
./infrastructure/deploy.sh --data      # Data files only
./infrastructure/deploy.sh --status    # Check server

See: .opencode/DEPLOYMENT_RULES.md and .opencode/MONOREPO_FRONTEND_APPS.md

Rule 8: Legal Form Terms MUST Be Filtered from CustodianName

🚨 CRITICAL: Exception to emic principle - Legal forms are ALWAYS filtered from CustodianName.

Examples: Stichting Rijksmuseum → CustodianName: Rijksmuseum, Legal Form: Stichting

Terms to Filter (by language):

Dutch: Stichting, B.V., N.V., Coöperatie
English: Foundation, Inc., Ltd., LLC, Corp.
German: Stiftung, e.V., GmbH, AG
French: Fondation, S.A., S.A.R.L.

NOT Filtered (part of identity): Vereniging, Association, Society, Verein

See: .opencode/LEGAL_FORM_FILTERING_RULE.md for complete documentation

Rule 9: Enum-to-Class Promotion - Single Source of Truth

🚨 CRITICAL: When an enum is promoted to a class hierarchy, the original enum MUST be deleted. Never maintain parallel enum/class definitions.

Archive Location: schemas/20251121/linkml/archive/enums/

See: .opencode/ENUM_TO_CLASS_PRINCIPLE.md for complete documentation

Rule 10: CH-Annotator is the Entity Annotation Convention

🚨 CRITICAL: All entity annotation follows ch_annotator-v1_7_0 convention.

9 Hypernym Types: AGT (Agent), GRP (Group), TOP (Toponym), GEO (Geometry), TMP (Temporal), APP (Appellation), ROL (Role), WRK (Work), QTY (Quantity)

Heritage Institutions: GRP.HER with GLAMORCUBESFIXPHDNT subtypes (GRP.HER.MUS, GRP.HER.LIB, GRP.HER.ARC, etc.)

See: .opencode/CH_ANNOTATOR_CONVENTION.md for complete documentation

Rule 11: Z.AI GLM API for LLM Tasks (NOT BigModel)

🚨 CRITICAL: Use Z.AI Coding Plan endpoint, NOT regular BigModel API.

Configuration:

API URL: https://api.z.ai/api/coding/paas/v4/chat/completions
Environment Variable: ZAI_API_TOKEN
Models: glm-4.5, glm-4.5-air, glm-4.5-flash, glm-4.6
Cost: Free (0 per token)

See: .opencode/ZAI_GLM_API_RULES.md for complete documentation

Rule 12: Person Data Reference Pattern - Avoid Inline Duplication

🚨 CRITICAL: Person profiles stored in data/custodian/person/entity/. Custodian files reference via person_profile_path - NEVER duplicate 50+ lines of profile data inline.

File Naming: {linkedin-slug}_{ISO-timestamp}.json

See: .opencode/PERSON_DATA_REFERENCE_PATTERN.md for complete documentation

Rule 13: Custodian Type Annotations on LinkML Schema Elements

🚨 CRITICAL: All schema elements MUST have custodian_types annotation with GLAMORCUBESFIXPHDNT single-letter codes.

Annotation Keys: custodian_types (list), custodian_types_rationale (string), custodian_types_primary (string)

Universal: Use ["*"] for elements applying to all types.

See: .opencode/CUSTODIAN_TYPE_ANNOTATION_CONVENTION.md for complete documentation

Rule 14: Exa MCP LinkedIn Profile Extraction

🚨 CRITICAL: Use exa_crawling_exa with direct URL for comprehensive LinkedIn profile extraction.

Tool Priority:

exa_crawling_exa - Profile URL known (preferred)
exa_linkedin_search_exa - Profile URL unknown
exa_web_search_exa - Fallback search

Output: data/custodian/person/entity/{linkedin-slug}_{timestamp}.json

See: .opencode/EXA_LINKEDIN_EXTRACTION_RULES.md for complete documentation

Rule 15: Connection Data Registration - Full Network Preservation

🚨 CRITICAL: ALL LinkedIn connections must be fully registered in dedicated connections files.

File Location: data/custodian/person/{slug}_connections_{timestamp}.json

Required: source_metadata, connections[] array, network_analysis with heritage type breakdown

See: .opencode/CONNECTION_DATA_REGISTRATION_RULE.md for complete documentation

Rule 16: LinkedIn Photo URLs - Store CDN URLs, Not Overlay Pages

🚨 CRITICAL: Store actual CDN URL, NOT overlay page URL.

❌ WRONG: linkedin.com/in/{slug}/overlay/photo/ (derivable, useless)
✅ CORRECT: media.licdn.com/dms/image/v2/{ID}/profile-displayphoto-shrink_800_800/...

See: .opencode/LINKEDIN_PHOTO_CDN_RULE.md for complete documentation

Rule 17: LinkedIn Connection Unique Identifiers

🚨 CRITICAL: Every connection gets unique ID including abbreviated and anonymous names.

Format: {target_slug}_conn_{index:04d}_{name_slug}

Name Types: full, abbreviated (Amy B.), anonymous (LinkedIn Member)

See: .opencode/LINKEDIN_CONNECTION_ID_RULE.md for complete documentation

Rule 18: Custodian Staff Parsing from LinkedIn Company Pages

🚨 CRITICAL: Use scripts/parse_custodian_staff.py for staff registration parsing.

Staff ID Format: {custodian_slug}_staff_{index:04d}_{name_slug}

See: .opencode/CUSTODIAN_STAFF_PARSING_RULE.md for complete documentation

Rule 19: HTML-Only LinkedIn Extraction (Preferred Method)

🚨 CRITICAL: Use ONLY manually saved HTML files for LinkedIn data extraction.

Data Completeness: HTML = 100% (including profile URLs), MD copy-paste = ~90%

Script: scripts/parse_linkedin_html.py

How to Save: Navigate → Scroll to load all → File > Save Page As > "Webpage, Complete"

See: .opencode/HTML_ONLY_LINKEDIN_EXTRACTION_RULE.md for complete documentation

Rule 20: Person Entity Profiles - Individual File Storage

🚨 CRITICAL: Person profiles stored as individual files in data/custodian/person/entity/.

File Naming: {linkedin-slug}_{ISO-timestamp}.json

Required: ALL profiles MUST use structured JSON with extraction_agent: "claude-opus-4.5". Raw content dumps are NOT acceptable.

See: .opencode/PERSON_ENTITY_PROFILE_FORMAT_RULE.md for complete documentation

Rule 21: Data Fabrication is Strictly Prohibited

🚨 CRITICAL: ALL DATA MUST BE REAL AND VERIFIABLE. Fabricating any data is strictly prohibited.

❌ FORBIDDEN:

Creating fake names, job titles, companies
Inventing education history or skills
Generating placeholder data when extraction fails
Creating fictional LinkedIn URLs

✅ ALLOWED:

Skip profiles that cannot be extracted
Return null or empty fields for missing data
Mark profiles with extraction_error: true
Log why extraction failed

See: .opencode/DATA_FABRICATION_PROHIBITION.md for complete documentation

Rule 22: Custodian YAML Files Are the Single Source of Truth

🚨 CRITICAL: data/custodian/*.yaml is the SINGLE SOURCE OF TRUTH for all enrichment data.

Data Hierarchy:

data/custodian/*.yaml  ← SINGLE SOURCE OF TRUTH
       ↓
Ducklake → PostgreSQL → TypeDB → Oxigraph → Qdrant
(All databases are DERIVED - never add data independently)
       ↓
REST API → Frontend (both DERIVED)

Workflow: FETCH → VALIDATE → WRITE TO YAML → Import to database → Verify

See: .opencode/CUSTODIAN_DATA_SOURCE_OF_TRUTH.md for complete documentation

🚨 CRITICAL: Social media links MUST be institution-specific, NOT generic platform homepages.

Invalid: facebook.com/, facebook.com/facebook, twitter.com/twitter

Valid: facebook.com/rijksmuseum/, twitter.com/rijksmuseum

See: .opencode/SOCIAL_MEDIA_LINK_VALIDATION.md for complete documentation

Rule 24: Unused Import Investigation - Check Before Removing

🚨 CRITICAL: Before removing unused imports, INVESTIGATE whether they indicate incomplete implementations.

Checklist:

Was it recently used? (git log -p --all -S 'ImportName')
Is there a TODO/FIXME?
Pattern mismatch (old vs new syntax)?
Incomplete feature?
Conditional usage (TYPE_CHECKING blocks)?

See: .opencode/UNUSED_IMPORT_INVESTIGATION_RULE.md for complete documentation

Rule 25: Digital Platform Discovery Enrichment

🚨 CRITICAL: Every heritage custodian MUST be enriched with digital platform discovery data.

Discover: Collection management systems, discovery portals, external integrations, APIs

Required Provenance: retrieval_agent, retrieval_timestamp, source_url, xpath_base, html_file

See: .opencode/DIGITAL_PLATFORM_DISCOVERY_RULE.md for complete documentation

Rule 26: Person Data Provenance - Web Claims for Staff Information

🚨 CRITICAL: All person/staff data MUST have web claim provenance with verifiable sources.

Required Fields: claim_type, claim_value, source_url, retrieved_on, retrieval_agent

Recommended: xpath, xpath_match_score

See: .opencode/PERSON_DATA_PROVENANCE_RULE.md for complete documentation

Rule 27: Person-Custodian Data Architecture

🚨 CRITICAL: Person entity files are the SINGLE SOURCE OF TRUTH for all person data.

In Person Entity File: extraction_metadata, profile_data, web_claims, affiliations

In Custodian YAML: person_id, person_name, role_title, affiliation_provenance, linkedin_profile_path (reference only)

NEVER: Put web_claims in custodian YAML files

See: .opencode/PERSON_CUSTODIAN_DATA_ARCHITECTURE.md for complete documentation

Rule 28: Web Claims Deduplication - No Redundant Claims

🚨 CRITICAL: Do not duplicate claims unless genuine variation exists with uncertainty.

Eliminate: Favicon variants, same value from different extractions, dynamic content

Document: Removed claims in removed_claims section for audit trail

See: .opencode/WEB_CLAIMS_DEDUPLICATION_RULE.md for complete documentation

Rule 29: Anonymous Profile Name Derivation from LinkedIn Slugs

🚨 CRITICAL: Names CAN be derived from hyphenated LinkedIn slugs - this is data transformation, NOT fabrication.

Dutch Particles: Keep lowercase when not first word (van, de, den, der)

Known Compound Slugs: Use mapping for jponjee → "J. Ponjee", etc.

See: .opencode/ANONYMOUS_PROFILE_NAME_RULE.md for complete documentation

Rule 30: Person Profile Extraction Confidence Scoring

🚨 CRITICAL: Every enriched profile MUST have confidence score (0.50-0.95) for data extraction quality.

Distinct from: Heritage sector relevance score (different purpose)

Scoring Factors:

Clear job title: +0.10 to +0.15
Named institution: +0.05 to +0.10
Privacy-abbreviated name: -0.15 to -0.20
Intern/trainee: -0.10

See: .opencode/PERSON_PROFILE_CONFIDENCE_SCORING.md for complete documentation

Rule 31: Organizational Subdivision Extraction

🚨 CRITICAL: ALWAYS capture organizational subdivisions as structured data.

Types: department, team, unit, division, section, lab_or_center, office

Store in: affiliations[].subdivision with type, name, parent_subdivision, extraction_source

See: .opencode/ORGANIZATIONAL_SUBDIVISION_EXTRACTION.md for complete documentation

Rule 32: Government Ministries Are Heritage Custodians (Type O)

🚨 CRITICAL: Government ministries ARE heritage custodians due to statutory record-keeping obligations.

Heritage Relevance Scores:

Role Category	Score Range
Records Management	0.40-0.50
IT/Systems (records)	0.30-0.40
Policy/Advisory	0.25-0.35
Administrative	0.15-0.25

See: .opencode/GOVERNMENT_MINISTRY_HERITAGE_RULE.md for complete documentation

Rule 33: GHCID Collision Duplicate Detection

🚨 CRITICAL: Duplicate detection is MANDATORY in GHCID collision resolution.

Decision Matrix:

ALL details match → DUPLICATE (keep earliest, archive later)
Same name, different city → NOT DUPLICATE (keep both, add suffix)
Same name, same city, different Wikidata IDs → NOT DUPLICATE
When in doubt → Keep both files (can merge later)

See: .opencode/GHCID_COLLISION_DUPLICATE_DETECTION.md for complete documentation

Rule 34: Linkup is the Preferred Web Scraper

🚨 CRITICAL: Use Linkup as primary web scraper. Firecrawl credits are limited.

Tool Priority:

Priority	Tool	When to Use
1st	`linkup_linkup-search`	General research, finding pages
2nd	`linkup_linkup-fetch`	Fetching known URL
3rd	`firecrawl_*`	Only when Linkup fails
4th	`playwright_*`	Interactive pages, HTML archival

Two-Phase for XPath Provenance (Rule 6 compliance):

Linkup for discovery
Playwright for archival with XPath extraction

See: .opencode/LINKUP_PREFERRED_WEB_SCRAPER_RULE.md for complete documentation

Rule 35: Provenance Statements MUST Have Dual Timestamps

🚨 CRITICAL: Every provenance statement MUST include at least TWO timestamps to distinguish when the claim was created from when the source was archived.

MANDATORY Timestamps:

Timestamp	Purpose	Example
`statement_created_at`	When the claim/annotation was extracted/created	`2025-12-30T14:30:00Z`
`source_archived_at`	When the source material was archived/captured	`2025-12-29T10:15:00Z`

Optional (Encouraged):

source_created_at - When the original source content was published
source_last_modified_at - When the source content was last updated
last_verified_at - When the claim was last re-verified
next_verification_due - When the claim should be re-verified

Example - CORRECT (Dual Timestamps):

provenance:
  statement_created_at: "2025-12-30T14:30:00Z"  # When we extracted this claim
  source_archived_at: "2025-12-29T10:15:00Z"   # When we archived the webpage
  source_created_at: "2022-07-15T00:00:00Z"    # Optional: article publish date

Example - WRONG (Single Timestamp):

# INVALID - Only one timestamp, vague agent
extraction_provenance:
  timestamp: '2025-11-06T08:02:44Z'  # Which timestamp is this?!
  agent: claude-conversation          # Too vague - which model?

Agent Identifier Standards:

❌ Invalid	✅ Valid
`claude-conversation`	`opencode-claude-sonnet-4`
`claude`	`opencode-claude-opus-4`
`ai`	`batch-script-python-3.11`
`opencode`	`manual-human-curator`

Validation Rule: source_archived_at MUST be ≤ statement_created_at (source archived before/when statement created)

Migration Note: 24,328 files in data/custodian/ with agent: claude-conversation require migration to dual timestamp format.

See: .opencode/PROVENANCE_TIMESTAMP_RULES.md for complete documentation

Rule 36: Original Language Preservation in Web Content Extraction

🚨 CRITICAL: ALL extracted text content MUST be preserved in its original source language. Translation is STRICTLY FORBIDDEN during extraction.

Applies to:

Mission statements
Vision statements
Organizational descriptions
About us content
Historical narratives
Collection descriptions
Any textual content extracted from institutional websites

Rationale:

Emic Authenticity - The institution's own voice and terminology must be preserved
Semantic Fidelity - Translation introduces interpretation and potential distortion
Provenance Integrity - Translated content breaks XPath provenance and content hash verification
Downstream Flexibility - Original content allows users to request translations in their preferred language

Required Fields:

mission_statement:
  text: "Het Rijksmuseum is het museum van Nederland..."  # Original Dutch
  language: "nl"  # ISO 639-1 code
  source_url: "https://www.rijksmuseum.nl/nl/over-ons"
  extracted_verbatim: true  # Confirms no translation occurred

LLM Prompt Requirements:

All LLM prompts for content extraction MUST include explicit no-translation instructions:

CRITICAL: Extract the text EXACTLY as it appears on the webpage.
DO NOT TRANSLATE. Preserve the original language.
If the source is in Dutch, the output must be in Dutch.
If the source is in Spanish, the output must be in Spanish.

Anti-Patterns (FORBIDDEN):

Scenario	Status
Translate Dutch → English during extraction	❌ FORBIDDEN
Store English text with Dutch `source_url`	❌ FORBIDDEN
Mix languages in extracted content	❌ FORBIDDEN
Omit `language` field	❌ FORBIDDEN

Validation Checklist:

Text is in original source language
language field matches content language
language matches expected from GHCID (or mismatch documented)
extracted_verbatim: true is set

See: .opencode/ORIGINAL_LANGUAGE_PRESERVATION_RULE.md for complete documentation including language-specific LLM prompts

Rule 37: Specificity Score Annotations for LinkML Classes

🚨 CRITICAL: Every LinkML class MUST have specificity score annotations to enable intelligent RAG retrieval filtering and UML visualization.

Annotation Format:

classes:
  ClassName:
    annotations:
      specificity_score: 0.75              # Required: 0.0-1.0
      specificity_rationale: "..."         # Required: Why this score
      template_specificity:                # Optional: Template-specific scores
        archive_search: 0.95
        museum_search: 0.20

Score Semantics (LOWER = more broadly relevant):

Score Range	Meaning	Examples
0.00-0.20	Universal	`HeritageCustodian`, `Location`
0.20-0.40	Broadly useful	`Collection`, `Identifier`
0.40-0.60	Moderately specific	`ChangeEvent`, `PersonProfile`
0.60-0.80	Fairly specific	`Archive`, `Museum`, `Library`
0.80-1.00	Highly specific	`LinkedInConnectionExtraction`

10 Conversation Templates for template_specificity:

archive_search, museum_search, library_search
collection_discovery, person_research, location_browse
identifier_lookup, organizational_change, digital_platform
general_heritage (fallback - uses specificity_score directly)

Validation Rules:

Score must be in range [0.0, 1.0]
Rationale must not be empty
Child class score must be ≥ parent class score (inheritance consistency)

Use Cases:

RAG Retrieval: Filter schema classes by relevance to user query
UML Visualization: Generate focused diagrams showing only relevant classes
Context Management: Reduce token usage by excluding low-relevance classes

See: .opencode/rules/specificity-score-convention.md for complete documentation

Rule 38: Slot Centralization and Semantic URI Requirements

🚨 CRITICAL: All LinkML slots MUST be centralized in schemas/20251121/linkml/modules/slots/ and MUST have semantically sound slot_uri predicates from base ontologies.

Key Requirements:

Centralization: All slots MUST be defined in modules/slots/, never inline in class files
slot_uri: Every slot MUST have a slot_uri from base ontologies (data/ontology/)
Mappings: Use exact_mappings, close_mappings, related_mappings, narrow_mappings, broad_mappings for additional semantic relationships

Why This Matters:

Frontend UML visualization depends on centralized slots for edge rendering
Semantic URIs enable linked data interoperability and RDF serialization
Mapping annotations connect to SKOS-based vocabulary alignment standards

Common slot_uri Sources:

Ontology	Prefix	Example Predicates
SKOS	`skos:`	`prefLabel`, `altLabel`, `definition`, `note`
Schema.org	`schema:`	`name`, `description`, `url`, `dateCreated`
Dublin Core	`dcterms:`	`identifier`, `title`, `creator`, `date`
PROV-O	`prov:`	`wasGeneratedBy`, `wasAttributedTo`, `atTime`
RiC-O	`rico:`	`hasRecordSetType`, `isOrWasPartOf`
CIDOC-CRM	`crm:`	`P1_is_identified_by`, `P2_has_type`

Workflow for New Slots:

Search data/ontology/ for existing predicate
Create file in modules/slots/ with slot_uri
Add mappings to related predicates in other ontologies
Update manifest.json with new slot file

See: .opencode/rules/slot-centralization-and-semantic-uri-rule.md for complete documentation

Rule 39: Slot Naming Convention (RiC-O Style)

🚨 CRITICAL: LinkML slots representing relational predicates MUST follow RiC-O-style naming conventions to express temporal semantics accurately.

Core Naming Patterns:

Pattern	Use Case	Examples
`hasOrHad*`	Temporal relationship (active voice)	`hasOrHadHolder`, `hasOrHadPart`, `hasOrHadType`
`isOrWas*`	Temporal relationship (inverse)	`isOrWasPartOf`, `isOrWasMemberOf`, `isOrWasHolderOf`
`has*`	Permanent/immutable facts	`hasBeginningDate`, `hasBirthPlace`, `hasIdentifier`
`*Transitive`	Hierarchical (through chain)	`isIncludedInTransitive`
`directly*`	Hierarchical (immediate only)	`directlyIncludes`, `isDirectlyIncludedIn`

Semantic Distinction: Hierarchy vs Association:

🚨 The same slot name can mask different semantics. Always analyze intent:

Category	Semantic	Pattern	Ontology
Organizational Hierarchy	"This org is part of that org"	RiC-O `isOrWas` / `hasOrHad`	RiC-O
Event Association	"This event happened to that entity"	PROV-O `wasAssociatedWith`	PROV-O

Example: The deprecated parent_custodian was used for TWO different semantics:

CustodianLegalStatus.parent_custodian → Hierarchy → Now: is_or_was_suborganization_of
OrganizationalChangeEvent.parent_custodian → Event association → Now: associated_custodian

Migration Mapping (Key Slots):

Deprecated	Replacement	Pattern
`parent_custodian` (hierarchy)	`is_or_was_suborganization_of`	RiC-O
`parent_custodian` (event)	`associated_custodian`	PROV-O
`has_suborganization`	`has_or_had_suborganization`	RiC-O
`parent_collection`	`is_or_was_sub_collection_of`	RiC-O
`sub_collections`	`has_or_had_sub_collection`	RiC-O
`has_collection`	`has_or_had_collection`	RiC-O
`encompassing_body`	`is_or_was_encompassed_by`	RiC-O
`has_member`	`has_or_had_member`	RiC-O
`is_member_of`	`is_or_was_member_of`	RiC-O

Decision Tree:

Is relationship about organizational structure?
├─ YES (child → parent) → isOrWas{Relationship}Of
├─ YES (parent → children) → hasOrHad{Relationship}
└─ NO (event → entity affected) → associated_custodian (PROV-O)

LinkML Slot Naming: Convert RiC-O predicates to snake_case:

rico:hasOrHadPart → has_or_had_part
rico:isOrWasPartOf → is_or_was_part_of

See: .opencode/rules/slot-naming-convention-rico-style.md for complete documentation

Rule 40: KIEN Registry is Authoritative for Intangible Heritage Custodians

🚨 CRITICAL: For Intangible Heritage Custodians (Type I), the KIEN registry at https://www.immaterieelerfgoed.nl/ is TIER_1_AUTHORITATIVE. Google Maps is TIER_3_CROWD_SOURCED and frequently returns false matches.

Why Google Maps Fails for Type I:

Virtual organizations without commercial storefronts
Name collisions with unrelated businesses (e.g., "Platform" → "Platform 9 BV")
No physical Google Maps presence for intangible heritage networks
Volunteer-run organizations with residential addresses

Data Tier Hierarchy for Type I:

Priority	Source	Data Tier
1st	KIEN Registry (`immaterieelerfgoed.nl`)	TIER_1_AUTHORITATIVE
2nd	Organization's Official Website	TIER_2_VERIFIED
3rd	Wikidata	TIER_3_CROWD_SOURCED
4th	Google Maps	TIER_3_CROWD_SOURCED (verify!)

Required Workflow:

Scrape KIEN page first - Extract address from Contact section
Validate Google Maps - Compare domain/name against KIEN data
Mark false matches - Set status: FALSE_MATCH with documentation

Marking False Matches:

google_maps_enrichment:
  status: FALSE_MATCH
  false_match_reason: "Google Maps returned different organization"
  original_false_match:
    place_id: ChIJ...
    name: "Wrong Business Name"
    website: "http://wrong-domain.nl/"
  correction_timestamp: "2025-01-08T00:00:00Z"

Location Resolution: Use KIEN address → Geocode with Nominatim → NOT Google Maps coordinates

See: .opencode/rules/kien-authoritative-source-rule.md for complete documentation

Rule 41: LinkML "Types" Classes Define SPARQL Template Variables

🚨 CRITICAL: LinkML classes following the *Type / *Types naming pattern (Rule 0b) serve as the single source of truth for valid values in SPARQL template slot variables.

When designing SPARQL templates, extract variables from the schema rather than hardcoding separate templates for each institution type or geographic level.

Why This Matters:

Same template works across ALL institution types (musea, archieven, bibliotheken, etc.)
Same template works across ALL geographic levels (country, subregion, settlement)
Adding new types to schema automatically extends template capabilities
Multilingual support comes for free from schema labels

Template Variable Sources:

Variable	Schema Source	Examples
`institution_type`	`CustodianType` + 19 subclasses	M (Museum), A (Archive), L (Library)
`location`	Hierarchical: Country/Subregion/Settlement	NL, NL-NH, Amsterdam
`platform_type`	`DigitalPlatformTypes.yaml` (69+ types)	DigitalLibrary, Aggregator

Template Design Pattern:

# CORRECT: Single parameterized template
count_institutions_by_type_location:
  slots:
    institution_type:
      schema_source: "modules/classes/CustodianType.yaml"
    location:
      resolution_order: [settlement, subregion, country]
      
  # SlotExtractor detects level and selects appropriate SPARQL variant
  sparql_template: |
    SELECT (COUNT(?s) AS ?count) WHERE {
      ?s hc:institutionType "{{ institution_type }}" ;
         hc:settlementName "{{ location }}" .
    }    
  sparql_template_region: |
    SELECT (COUNT(?s) AS ?count) WHERE {
      ?s hc:institutionType "{{ institution_type }}" ;
         hc:subregionCode "{{ location }}" .
    }

SlotExtractor Responsibilities:

Detect institution type from query: "musea" → M, "archieven" → A
Detect location level: "Amsterdam" → settlement, "Noord-Holland" → subregion
Normalize values: "Noord-Holland" → "NL-NH"

See: .opencode/rules/types-classes-as-template-variables.md for complete documentation

Rule 42: No Ontology Prefixes in Slot Names

🚨 CRITICAL: LinkML slot names MUST NOT include ontology namespace prefixes. Ontology references belong in mapping properties (slot_uri, exact_mappings, close_mappings, etc.), NOT in element names.

Why This Matters:

Slot names should be human-readable, domain-focused terminology
Ontology mappings are documented via LinkML's dedicated mapping properties
Embedding prefixes creates coupling between naming and specific ontology versions
Clean separation allows renaming slots without changing ontology bindings

Prohibited Prefixes:

Prefix	Ontology	Example Violation
`rico_`	Records in Contexts	`rico_organizational_principle`
`skos_`	SKOS	`skos_broader`, `skos_narrower`
`schema_`	Schema.org	`schema_name`
`dcterms_`	Dublin Core	`dcterms_created`
`prov_`	PROV-O	`prov_generated_by`
`org_`	W3C Organization	`org_has_member`
`crm_`	CIDOC-CRM	`crm_carried_out_by`
`foaf_`	FOAF	`foaf_knows`

Correct Pattern:

# CORRECT: Clean name with ontology reference in slot_uri and mappings
slots:
  record_holder:
    description: The custodian that holds or held this record set.
    slot_uri: rico:hasOrHadHolder
    exact_mappings:
      - rico:hasOrHadHolder
    close_mappings:
      - schema:holdingArchive
    range: Custodian

WRONG Pattern:

# WRONG: Ontology prefix embedded in slot name
slots:
  rico_has_or_had_holder:  # BAD - "rico_" prefix duplicates slot_uri info
    slot_uri: rico:hasOrHadHolder
    range: string

Exceptions:

External identifier slots: wikidata_id, viaf_id, isil_code (system names, not ontology prefixes)
Internal technical slots: internal_wd_namespace_force (prefixed with internal_)

See: .opencode/rules/no-ontology-prefix-in-slot-names.md for complete documentation and migration examples

Rule 43: Slot Nouns Must Be Singular

🚨 CRITICAL: LinkML slot names MUST use singular nouns, even for multivalued slots. The multivalued: true property indicates cardinality, NOT the slot name.

Rationale:

Predicate semantics: Slots represent predicates/relationships. In RDF, hasCollection can have multiple objects without changing the predicate name.
Consistency: Singular names work for both single-valued and multivalued slots.
Ontology alignment: Standard ontologies use singular predicates (skos:broader, org:hasMember, rico:hasOrHadHolder).
Readability: custodian.has_or_had_custodian_type reads naturally as "custodian has (or had) custodian type".

Correct Pattern:

slots:
  has_or_had_custodian_type:        # ✅ CORRECT - singular noun
    slot_uri: org:classification
    range: CustodianType
    multivalued: true               # Cardinality expressed here, not in name

  has_or_had_collection:            # ✅ CORRECT - singular noun
    range: CustodianCollection
    multivalued: true
    
  has_or_had_member:                # ✅ CORRECT - singular noun
    range: Custodian
    multivalued: true

WRONG Pattern:

slots:
  has_or_had_custodian_types:       # ❌ WRONG - plural noun
    multivalued: true
    
  collections:                       # ❌ WRONG - plural noun
    multivalued: true

Migration Examples:

Old (Plural)	New (Singular)
`custodian_types`	`has_or_had_custodian_type`
`collections`	`has_or_had_collection`
`identifiers`	`identifier`
`alternative_names`	`alternative_name`
`staff_members`	`staff_member`

Exceptions (compound concepts where plural is part of the proper noun):

archives_regionales - French administrative term
united_states - Geographic proper noun

See: .opencode/rules/slot-noun-singular-convention.md for complete documentation

Rule 44: PPID Birth Date Enrichment and EDTF Unknown Date Notation

🚨 CRITICAL: When birth/death dates are missing from person entity sources, agents MUST first attempt enrichment via web search. Only after comprehensive search fails should EDTF unknown notation be used.

Enrichment Workflow:

Search Exa: "{full_name}" born birthday birth date
Search Linkup: "{name}" biography
If found → Record as web_claim with provenance
If NOT found → Use EDTF notation with enrichment_metadata recording the failed search

EDTF Notation (Library of Congress Standard):

Character	Meaning	Example
`X`	Unspecified digit	`197X` = some year 1970-1979
`~`	Approximate (circa)	`1985~` = circa 1985
`?`	Uncertain	`1985?` = possibly 1985
`S`	Significant digits	`1975S3` = estimated 1975, accurate to decade
`[..]`	One of set	`[197X,198X]` = 1970s or 1980s

Common Patterns:

Scenario	EDTF Format
Decade known (1970s)	`197X`
Century known (1900s)	`19XX`
Completely unknown	`XXXX`
Multiple possible decades	`[197X,198X]`
Estimated from career	`1975S3`

Filename Safety: PPID filenames must avoid ?, %, [], /, |. Use simplified form in filename, full EDTF in metadata.

Anti-Patterns:

❌ "1970s" - Use 197X instead
❌ "circa 1985" - Use 1985~ instead
❌ "unknown" - Use XXXX instead
❌ Custom notation like 197~8~ - Not EDTF compliant

Validation:

Cannot use XXXX without enrichment_metadata.birth_date_search.attempted: true
All dates must parse as valid EDTF

See: .opencode/rules/ppid-birth-date-enrichment-rule.md for complete documentation

Rule 45: Inferred Data Must Be Explicit with Provenance

🚨 CRITICAL: All inferred data MUST be stored in explicit inferred_* fields with full provenance statements. Inferred values MUST NEVER silently replace or merge with verified data.

Required Inferred Fields (for person profiles):

Inferred Field	Source Observations	Heuristic
`inferred_birth_decade`	Earliest education/job dates	Entry age assumptions
`inferred_birth_settlement`	School/university location	Residential proximity
`inferred_current_settlement`	Profile location, current job	Direct extraction

Required Structure:

{
  "inferred_birth_decade": {
    "value": "196X",
    "edtf": "196X",
    "confidence": "low",
    "inference_provenance": {
      "method": "earliest_education_heuristic",
      "inference_chain": [
        {"step": 1, "observation": "University start 1986", "source_field": "profile_data.education[0]"},
        {"step": 2, "assumption": "University entry at age 18", "rationale": "Dutch standard"},
        {"step": 3, "calculation": "1986 - 18 = 1968", "result": "Birth year ~1968"},
        {"step": 4, "generalization": "Round to decade", "result": "196X"}
      ],
      "inferred_at": "2025-01-09T18:00:00Z",
      "inferred_by": "enrich_ppids.py"
    }
  }
}

Anti-Patterns:

❌ Silent replacement: Putting inferred value directly in birth_date.edtf without marking it as inferred
❌ Hidden metadata: Separating inference flag from the value
❌ Missing chain: Not documenting HOW the value was derived

PPID Component Tracking:

{
  "ppid_components": {
    "first_date": "196X",
    "first_date_source": "inferred_birth_decade",
    "first_location": "NL-UT-UTR",
    "first_location_source": "inferred_birth_settlement"
  }
}

List-Valued Inferred Data (EDTF Set Notation):

When inference yields multiple plausible values (e.g., decade boundary cases), store as a list:

{
  "inferred_birth_decade": {
    "values": ["196X", "197X"],
    "edtf": "[196X,197X]",
    "primary_value": "196X",
    "primary_rationale": "1965 is in 196X, but range extends into 197X",
    "confidence": "very_low"
  }
}

For PPID generation, use primary_value:

first_date_source: "inferred_birth_decade.primary_value"
first_date_alternatives: ["197X"]

See: .opencode/rules/inferred-data-explicit-provenance-rule.md for complete documentation

Rule 46: Entity Resolution - Names Are NEVER Sufficient

🚨 CRITICAL: Similar or identical names are NEVER sufficient for entity resolution. When enriching person profiles via web search, verify MULTIPLE identity attributes before attributing ANY claim.

⚠️ ALL ENRICHMENTS MUST BE DOUBLE-CHECKED ⚠️

Wrong data is worse than no data. A single false claim (birth year from wrong person, spouse from a different namesake, social media from random account) corrupts the entire profile and undermines dataset trustworthiness.

The Core Problem:

Web searches for "Carmen Juliá" return data about Carmen Julia Álvarez (Venezuelan actress), Carmen Julia Navarro (Mexican hydrogeologist), and Carmen Julia Gutiérrez (Spanish medievalist)
None of these is the actual Carmen Juliá who is a UK art curator
Name matching alone caused 122+ false birth year attributions

Required Identity Attributes (need 3 of 5 to match):

#	Attribute	Check	Conflict Example
1	Career/Profession	Same field	Source: "actress", Profile: "curator" → REJECT
2	Employer	Same institution	Source: "film studio", Profile: "museum" → REJECT
3	Location	Same city/country	Source: "Venezuela", Profile: "UK" → REJECT
4	Age Range	Plausible for career	Source: "born 1922", Profile: active 2025 → REJECT
5	Education	Same field	Source: "medical school", Profile: "art history" → REJECT

High-Risk Sources (require stricter verification, NOT forbidden):

Genealogy sites (geni.com, ancestry., myheritage.) → Require 5/5 matches
IMDB (actors with same name) → Require 5/5 matches (unless person works in film/TV)
Wikipedia articles → Require 4/5 matches
Social media → Require 4/5 matches + verify employer/location in bio

Red Flags Requiring Investigation (not automatic rejection):

Profession difference (actress vs curator) → Investigate: did they change careers?
Location difference (Venezuela vs UK) → Investigate: did they relocate?
Time gap in career → Investigate: career break or different person?

When to REJECT: If investigation shows no plausible connection (e.g., Venezuelan actress active 1970s-2000s cannot be UK curator active 2020s - overlapping timelines, different continents)

Name Commonality Thresholds:

Name Type	Required Matches
Unique (e.g., "Xander Vermeulen-Oosterhuis")	2 of 5
Moderate (e.g., "Carmen Juliá")	3 of 5
Common (e.g., "Jan de Vries")	4 of 5
Very common (e.g., "John Smith")	5 of 5 or reject

See: .opencode/rules/entity-resolution-no-heuristics.md for complete documentation

Rule 47: Disambiguation Entity Profiles - Prevent Repeated Errors

🚨 CRITICAL: When entity resolution determines that a web source describes a different person with a similar name, create a PPID profile for that person. The PPID system is universal - ANY person who ever lived can have a profile.

The Universal PPID Principle:

ALL persons on Earth can be assigned PPIDs - heritage workers and non-heritage alike
Historical persons (deceased from any era) can have profiles
The heritage_relevance field indicates heritage sector relevance, NOT profile eligibility
Anyone can have a PPID - the actress, the doctor, the footballer

Why This Matters:

Prevents future enrichment from making the same mistake
Documents investigation work
Builds comprehensive person database
Bidirectional linking between profiles

When to Create Namesake Profiles:

Entity resolution rejects a claim due to identity mismatch
The namesake is notable enough to appear in search results repeatedly
The confusion risk is high (similar name, some overlapping attributes)

Example: Carmen Julia Álvarez (Venezuelan actress)

Discovered during enrichment of Carmen Juliá (UK curator)
Different profession (actress vs curator), different location (Venezuela vs UK)
Create regular PPID: ID_VE-XX-CCS_1952_VE-XX-CCS_XXXX_CARMEN-JULIA-ALVAREZ.json
Set heritage_relevance.is_heritage_relevant: false
Link both profiles via disambiguation_notes

Namesake Profile Structure:

{
  "ppid": "ID_VE-XX-CCS_1952_VE-XX-CCS_XXXX_CARMEN-JULIA-ALVAREZ",
  "profile_data": {
    "full_name": "Carmen Julia Álvarez",
    "profession": "actress",
    "birth_year": 1952
  },
  "heritage_relevance": {
    "is_heritage_relevant": false,
    "relevance_score": 0.0
  },
  "disambiguation_notes": {
    "commonly_confused_with": [
      {"ppid": "ID_UK-XX-XXX_...", "name": "Carmen Juliá", "profession": "curator"}
    ]
  }
}

See: .opencode/rules/disambiguation-entity-profiles.md for complete documentation

Appendix: Full Rule Content (No .opencode Equivalent)

The following rules have no separate .opencode file and are preserved in full:

Rule 2: Wikidata Entities Are NOT Ontology Classes

Files:

data/wikidata/GLAMORCUBEPSXHFN/hyponyms_curated.yaml
data/wikidata/GLAMORCUBEPSXHFN/hyponyms_curated_full.yaml

These files contain:

✅ Wikidata entity identifiers (Q-numbers) for heritage institution TYPES
✅ Multilingual labels and descriptions
✅ Hypernym classifications (upper-level categories)
✅ Source data for ontology mapping analysis

These files DO NOT contain:

❌ Formal ontology class definitions
❌ Direct class_uri mappings for LinkML
❌ Ontology properties or relationships

REQUIRED WORKFLOW:

hyponyms_curated.yaml (Wikidata Q-numbers)
    ↓
ANALYZE semantic meaning + hypernyms
    ↓
SEARCH base ontologies for matching classes
    ↓
MAP Wikidata entity → Ontology class(es)
    ↓
DOCUMENT rationale + properties
    ↓
CREATE LinkML schema with ontology class_uri

Example - WRONG ❌:

Mansion:
  class_uri: wd:Q1802963  # ← This is an ENTITY, not a CLASS!

Example - CORRECT ✅:

Mansion:
  # Wikidata source: Q1802963
  place_aspect:
    class_uri: crm:E27_Site  # CIDOC-CRM ontology class
  custodian_aspect:
    class_uri: cpov:PublicOrganisation  # If operates as museum

Rule 3: Multi-Aspect Modeling is Mandatory

Every heritage entity has MULTIPLE ontological aspects with INDEPENDENT temporal lifecycles.

Required Aspects:

Place Aspect (physical location/site)
- Ontology: crm:E27_Site + schema:Place
- Temporal: Construction → Demolition/Present
Custodian Aspect (organization managing heritage)
- Ontology: cpov:PublicOrganisation OR schema:Organization
- Temporal: Founding → Dissolution/Present
Legal Form Aspect (legal entity registration)
- Ontology: org:FormalOrganization + tooi:Overheidsorganisatie (Dutch)
- Temporal: Registration → Deregistration/Present
Collections Aspect (heritage materials)
- Ontology: rico:RecordSet OR crm:E78_Curated_Holding OR bf:Collection
- Temporal: Accession → Deaccession (per item)
People Aspect (staff, curators)
- Ontology: pico:PersonObservation + crm:E21_Person
- Temporal: Employment start → Employment end (per person)
Temporal Events (organizational changes)
- Ontology: crm:E10_Transfer_of_Custody, rico:Event
- Tracks custody transfers, mergers, relocations, transformations

Example: A historic mansion operating as a museum has:

Place aspect: Building constructed 1880, still standing (143 years)
Custodian aspect: Foundation established 1994 to operate museum (30 years)
Legal form: Dutch stichting registered 1994, KvK #12345678
Collections: Mondrian artworks acquired 1994-2024
People: Current curator employed 2020-present

Each aspect changes independently over time!

Rule 5: NEVER Delete Enriched Data - Additive Only

🚨 CRITICAL: Data enrichment is ADDITIVE ONLY. Never delete or overwrite existing enriched content.

When restructuring or updating enriched institution records:

✅ ALLOWED (Additive Operations):

Add new fields or sections
Restructure YAML/JSON layout while preserving all content
Rename files (e.g., _unknown.yaml → _museum_name.yaml)
Add provenance metadata
Merge data from multiple sources (preserving all)

❌ FORBIDDEN (Destructive Operations):

Delete Google Maps data (reviews, ratings, photo counts, popular times)
Remove OpenStreetMap metadata
Overwrite website scrape results
Delete Wikidata enrichment data
Remove any *_enrichment sections
Truncate or summarize detailed content

Data Types That Must NEVER Be Deleted:

Data Source	Protected Fields
Google Maps	`reviews`, `rating`, `total_ratings`, `photo_count`, `popular_times`, `place_id`, `business_status`
OpenStreetMap	`osm_id`, `osm_type`, `osm_tags`, `amenity`, `building`, `heritage`
Wikidata	`wikidata_id`, `claims`, `sitelinks`, `aliases`, `descriptions`
Website Scrape	`organization_details`, `collections`, `exhibitions`, `contact`, `social_media`, `accessibility`
ISIL Registry	`isil_code`, `assigned_date`, `remarks`

Example - CORRECT Restructuring:

# BEFORE (flat structure)
google_maps_rating: 4.5
google_maps_reviews: 127
website_description: "Historic museum..."

# AFTER (nested structure) - ALL DATA PRESERVED
enrichment_sources:
  google_maps:
    rating: 4.5          # ← PRESERVED
    reviews: 127         # ← PRESERVED
  website:
    description: "Historic museum..."  # ← PRESERVED

Example - WRONG (Data Loss):

# BEFORE
google_maps_enrichment:
  rating: 4.5
  reviews: 127
  popular_times: {...}
  photos: [...]

# AFTER - WRONG! Data deleted!
enrichment_status: enriched
# Where did the rating, reviews, popular_times go?!

Rationale:

Enriched data is expensive to collect (API calls, rate limits, web scraping)
Google Maps data changes over time - historical snapshots are valuable
Reviews and ratings provide quality signals for heritage institutions
Photo metadata enables visual discovery and verification
Deleting data violates data provenance principles

If Unsure: When restructuring files, first READ the entire file, then WRITE a new version that includes ALL original content in the new structure.

Rule 6: WebObservation Claims MUST Have XPath Provenance

Every claim extracted from a webpage MUST have an XPath pointer to the exact location in archived HTML where that value appears. Claims without XPath provenance are FABRICATED and must be removed.

This is not about "confidence" or "uncertainty" - it's about verifiability. Either the claim value exists in the HTML at a specific XPath, or it was hallucinated/fabricated by an LLM.

Required Fields for WebObservation Claims:

Field	Required	Description
`claim_type`	YES	Type of claim (full_name, description, email, etc.)
`claim_value`	YES	The extracted value
`source_url`	YES	URL the claim was extracted from
`retrieved_on`	YES	ISO 8601 timestamp when page was archived
`xpath`	YES	XPath to the element containing this value
`html_file`	YES	Relative path to archived HTML file
`xpath_match_score`	YES	1.0 for exact match, <1.0 for fuzzy match

Example - CORRECT (Verifiable):

web_enrichment:
  claims:
    - claim_type: full_name
      claim_value: Historische Vereniging Nijeveen
      source_url: https://historischeverenigingnijeveen.nl/
      retrieved_on: "2025-11-29T12:28:00Z"
      xpath: /[document][1]/html[1]/body[1]/div[6]/div[1]/table[3]/tbody[1]/tr[1]/td[1]/p[6]
      html_file: web/0021/historischeverenigingnijeveen.nl/rendered.html
      xpath_match_score: 1.0

Example - WRONG (Fabricated - Must Be Removed):

web_enrichment:
  claims:
    - claim_type: full_name
      claim_value: Historische Vereniging Nijeveen
      confidence: 0.95  # ← NO! This is meaningless without XPath

Workflow:

Archive website using Playwright: python scripts/fetch_website_playwright.py <entry> <url>
Add XPath provenance: python scripts/add_xpath_provenance.py
Script removes fabricated claims (stored in removed_unverified_claims for audit)

See:

.opencode/WEB_OBSERVATION_PROVENANCE_RULES.md for complete documentation
schemas/20251121/linkml/modules/classes/WebClaim.yaml for LinkML schema definition

Scope Clarification: This rule applies to WebClaim and WebObservation classes only. Other data classes have different provenance models:

CustodianTimelineEvent: Source-agnostic design - use extraction_notes for API queries/XPaths, and observation_ref to link to WebObservation/CustodianObservation for detailed provenance. See .opencode/PROVENANCE_SEPARATION_RULE.md.
GoogleMapsEnrichment: Uses Place ID and API response provenance.
WikidataEnrichment: Uses entity ID and SPARQL query provenance.

Project Overview

Goal: Extract structured data about worldwide GLAMORCUBESFIXPHDNT (Galleries, Libraries, Archives, Museums, Official institutions, Research centers, Corporations, Unknown, Botanical gardens/zoos, Educational providers, Societies, Features, Intangible heritage groups, miXed, Personal collections, Holy sites, Digital platforms, NGOs, Taste/smell heritage) institutions from 139+ Claude conversation JSON files and integrate with authoritative CSV datasets.

Output: Validated LinkML-compliant records representing heritage custodian organizations with provenance tracking, geographic data, identifiers, and relationship information.

Schema: See the modular LinkML schema v0.2.1 with 19-type GLAMORCUBESFIXPHDNT taxonomy described below.

Schema Reference (v0.2.1)

The project uses a modular LinkML schema organized into 6 specialized modules:

schemas/heritage_custodian.yaml - Main schema (import-only structure)
- Top-level schema that imports all modules
- Defines schema metadata and namespace
schemas/core.yaml - Core Classes
- HeritageCustodian - Main institution entity
- Location - Geographic data
- Identifier - External identifiers (ISIL, Wikidata, VIAF, etc.)
- DigitalPlatform - Online systems and platforms
- GHCID - Global Heritage Custodian Identifier
schemas/enums.yaml - Enumerations
- InstitutionTypeEnum - 13 institution types (GALLERY, LIBRARY, ARCHIVE, MUSEUM, etc.)
- ChangeTypeEnum - 11 organizational change types (FOUNDING, MERGER, CLOSURE, etc.)
- DataSource - Data origin types (CSV_REGISTRY, CONVERSATION_NLP, etc.)
- DataTier - Data quality tiers (TIER_1_AUTHORITATIVE through TIER_4_INFERRED)
- PlatformTypeEnum - Digital platform categories
schemas/provenance.yaml - Provenance Tracking
- Provenance - Data source and quality metadata
- ChangeEvent - Organizational change history (mergers, relocations, etc.)
- GHCIDHistoryEntry - GHCID change tracking over time
schemas/collections.yaml - Collection Metadata
- Collection - Collection descriptions
- Accession - Acquisition records
- DigitalObject - Digital surrogates
schemas/dutch.yaml - Dutch-Specific Extensions
- DutchHeritageCustodian - Netherlands heritage institutions
- Extensions for ISIL registry, platform integrations, KvK numbers

See /docs/SCHEMA_MODULES.md for detailed architecture and design patterns.

Base Ontologies for Global GLAM Data

CRITICAL: Before designing extraction pipelines or extending the schema, AI agents MUST consult the base ontologies that the LinkML schema builds upon. These ontologies provide standardized vocabularies and patterns for modeling heritage institutions.

Foundation Ontologies

The GLAM project integrates with three primary ontologies, each serving different geographic and semantic scopes:

1. TOOI - Dutch Government Organizational Ontology

File: /data/ontology/tooiont.ttl
Namespace: https://identifier.overheid.nl/tooi/def/ont/
Scope: Dutch heritage institutions (government archives, state museums, public cultural organizations)

When to Use:

✅ Extracting Dutch heritage institutions from conversations
✅ Modeling Dutch organizational change events (mergers, splits, reorganizations)
✅ Integrating with Dutch ISIL registry or KvK (Chamber of Commerce) data
✅ Parsing Dutch government heritage agency data

Key Classes:

tooi:Overheidsorganisatie - Government organization (extends to DutchHeritageCustodian)
tooi:Wijzigingsgebeurtenis - Change event (founding, merger, closure, relocation)

Key Properties:

tooi:officieleNaamInclSoort - Official name including type
tooi:begindatum / tooi:einddatum - Temporal validity (start/end dates)
tooi:organisatieIdentificatie - Formal identifiers (ISIL codes, etc.)

LinkML Mapping:

# schemas/dutch.yaml extends TOOI
DutchHeritageCustodian:
  is_a: HeritageCustodian
  class_uri: tooi:Overheidsorganisatie  # ← Maps to TOOI

Reference: See /docs/ONTOLOGY_EXTENSIONS.md for complete TOOI integration patterns.

2. CPOV - EU Core Public Organisation Vocabulary

Files:

/data/ontology/core-public-organisation-ap.ttl (RDF schema)
/data/ontology/core-public-organisation-ap.jsonld (JSON-LD context)

Namespace: http://data.europa.eu/m8g/
Scope: EU-wide and global public sector heritage organizations

When to Use:

✅ Extracting European heritage institutions (France, Germany, Belgium, etc.)
✅ Modeling international/global heritage organizations
✅ Aligning with EU Linked Open Data initiatives (Europeana, DPLA)
✅ Extracting non-Dutch institutions from conversations

Key Classes:

cpov:PublicOrganisation - Public sector organization (base for HeritageCustodian)
cv:ChangeEvent - Organizational change events
locn:Address - Physical location data

Key Properties:

skos:prefLabel / skos:altLabel - Preferred and alternative names
dct:identifier - Formal identifiers (ISIL, Wikidata, VIAF)
dct:temporal - Temporal coverage (founding to closure dates)
locn:address - Physical addresses

LinkML Mapping:

# schemas/core.yaml aligns with CPOV
HeritageCustodian:
  class_uri: cpov:PublicOrganisation  # ← Maps to CPOV
  
  slots:
    name:
      slot_uri: skos:prefLabel
    alternative_names:
      slot_uri: skos:altLabel
    identifiers:
      slot_uri: dct:identifier

Reference: See /docs/ONTOLOGY_EXTENSIONS.md for complete CPOV integration patterns.

3. Schema.org - Web Vocabulary for Structured Data

File: /data/ontology/schemaorg.owl
Namespace: http://schema.org/
Scope: Universal web semantics (museums, galleries, collections, events, learning resources)

When to Use:

✅ Extracting private collections or non-governmental organizations
✅ Modeling digital platforms (learning management systems, discovery portals)
✅ Web discoverability and SEO optimization
✅ Fallback when TOOI/CPOV don't apply

Key Classes:

schema:Museum / schema:Library / schema:ArchiveOrganization - Heritage institution types
schema:Place - Geographic locations
schema:LearningResource - Educational platforms (LMS, online courses)
schema:Event - Organizational events (founding, exhibitions)

LinkML Mapping:

# schemas/enums.yaml maps platform types to Schema.org
DigitalPlatformTypeEnum:
  LEARNING_MANAGEMENT:
    meaning: schema:LearningResource  # ← Maps to Schema.org

Reference: See /docs/ONTOLOGY_EXTENSIONS.md for Schema.org usage examples.

Ontology Decision Tree for Agents

When extracting heritage institution data, choose the appropriate ontology:

START: Extract institution from conversation
  ↓
Is the institution Dutch?
  ├─ YES → Use TOOI ontology
  │         - Map to schemas/dutch.yaml
  │         - Extract ISIL codes (NL-* format)
  │         - Extract KvK numbers (8-digit)
  │         - Model change events as tooi:Wijzigingsgebeurtenis
  │
  └─ NO → Is it a public/government organization?
           ├─ YES → Use CPOV ontology
           │         - Map to schemas/core.yaml
           │         - Extract standard identifiers (ISIL, Wikidata, VIAF)
           │         - Model change events as cv:ChangeEvent
           │
           └─ NO → Use Schema.org
                    - Map to schemas/core.yaml
                    - Use schema:Museum, schema:Library, etc.
                    - Emphasize web discoverability

Multi-Ontology Support: Institutions can implement MULTIPLE ontology classes simultaneously:

<https://w3id.org/heritage/custodian/nl/rijksmuseum>
    a tooi:Overheidsorganisatie,  # Dutch government organization
      cpov:PublicOrganisation,        # EU public sector
      schema:Museum ;                 # Schema.org web semantics

Required Ontology Consultation Workflow

Before extracting data, agents MUST perform these steps:

Step 1: Identify Institution Geographic Scope

# Determine which ontology applies
if institution_country == "NL":
    primary_ontology = "TOOI"
    ontology_file = "/data/ontology/tooiont.ttl"
elif institution_in_europe or institution_public_sector:
    primary_ontology = "CPOV"
    ontology_file = "/data/ontology/core-public-organisation-ap.ttl"
else:
    primary_ontology = "Schema.org"
    ontology_file = "/data/ontology/schemaorg.owl"

Step 2: Review Ontology Classes and Properties

Search ontology files for relevant classes:

# Dutch institutions - search TOOI
rg "tooi:Overheidsorganisatie|Wijzigingsgebeurtenis|begindatum" /data/ontology/tooiont.ttl

# EU/global institutions - search CPOV
rg "cpov:PublicOrganisation|cv:ChangeEvent|locn:Address" /data/ontology/core-public-organisation-ap.ttl

# All institutions - search Schema.org
rg "schema:Museum|schema:Library|schema:ArchiveOrganization" /data/ontology/schemaorg.owl

Step 3: Map Conversation Data to Ontology Properties

Create a mapping table before extraction:

Extracted Field	TOOI Property	CPOV Property	Schema.org Property
Institution name	`tooi:officieleNaamInclSoort`	`skos:prefLabel`	`schema:name`
Alternative names	-	`skos:altLabel`	`schema:alternateName`
Founding date	`tooi:begindatum`	`schema:startDate`	`schema:foundingDate`
Closure date	`tooi:einddatum`	`schema:endDate`	`schema:dissolutionDate`
ISIL code	`tooi:organisatieIdentificatie`	`dct:identifier`	`schema:identifier`
Address	(use `locn:Address`)	`locn:address`	`schema:address`
Merger event	`tooi:Wijzigingsgebeurtenis`	`cv:ChangeEvent`	`schema:Event`
Website	-	`schema:url`	`schema:url`

Step 4: Document Ontology Alignment in Provenance

Always include ontology references in extraction metadata:

provenance:
  data_source: CONVERSATION_NLP
  extraction_method: "NLP extraction following CPOV ontology patterns"
  base_ontology: "http://data.europa.eu/m8g/"  # ← Document which ontology used
  ontology_alignment:
    - "cpov:PublicOrganisation"
    - "cv:ChangeEvent"
  extraction_date: "2025-11-09T..."

Common Ontology Patterns

Pattern 1: Organizational Change Events

When extracting mergers, splits, relocations, name changes:

# TOOI pattern (Dutch institutions)
change_history:
  - event_id: https://w3id.org/heritage/custodian/event/nha-merger-2001
    change_type: MERGER  # Maps to tooi:Wijzigingsgebeurtenis
    event_date: "2001-01-01"
    event_description: "Merger of Gemeentearchief Haarlem and Rijksarchief in Noord-Holland"
    ontology_class: "tooi:Wijzigingsgebeurtenis"

# CPOV pattern (EU/global institutions)
change_history:
  - event_id: https://w3id.org/heritage/custodian/event/bnf-founding
    change_type: FOUNDING  # Maps to cv:ChangeEvent
    event_date: "1461-01-01"
    event_description: "Founded by King Louis XI as Royal Library"
    ontology_class: "cv:ChangeEvent"

Pattern 2: Multilingual Names

CPOV and Schema.org support language-tagged literals:

name: Bibliothèque nationale de France
alternative_names:
  - National Library of France@en
  - BnF@fr
  - Französische Nationalbibliothek@de

# RDF serialization:
# skos:prefLabel "Bibliothèque nationale de France"@fr ;
# skos:altLabel "National Library of France"@en, "BnF"@fr ;

Pattern 3: Hierarchical Relationships

Use W3C Org Ontology patterns (integrated in CPOV):

# Parent institution
parent_organization:
  name: Ministry of Culture
  relationship_type: "org:hasUnit"  # CPOV uses W3C Org Ontology
  
# Branch institutions
branches:
  - name: Regional Archive Noord-Brabant
    relationship_type: "org:subOrganizationOf"

Anti-Patterns to Avoid

❌ DON'T: Invent custom properties when ontology equivalents exist

# BAD - Custom property instead of ontology reuse
institution_official_name: "Rijksarchief"  # Use skos:prefLabel instead!

❌ DON'T: Ignore ontology namespace conventions

# BAD - No ontology reference
change_type: "merger"  # Use cv:ChangeEvent with proper namespace!

❌ DON'T: Extract without reviewing ontology files

# BAD - Extracting Dutch institutions without reading TOOI
agent: "I'll extract Dutch archives using Schema.org only"
# This loses semantic precision and ignores domain-specific patterns!

✅ DO: Always map to base ontologies and document alignment

# GOOD - Ontology-aligned extraction
name: Rijksarchief in Noord-Holland
institution_type: ARCHIVE
ontology_class: tooi:Overheidsorganisatie  # ← Documented
provenance:
  base_ontology: "https://identifier.overheid.nl/tooi/def/ont/"
  ontology_alignment:
    - tooi:Overheidsorganisatie
    - prov:Organization  # TOOI uses PROV-O for temporal tracking

Additional Ontology Resources

CIDOC-CRM (Cultural Heritage Domain):

File: /data/ontology/CIDOC_CRM_v7.1.3.rdf
Use for: Museum object cataloging, provenance, conservation
Key classes: crm:E74_Group (organizations), crm:E5_Event (historical events)

RiC-O (Records in Contexts - Archival Description):

Use for: Archival collections, fonds, series, items
Key classes: rico:CorporateBody, rico:RecordSet
Integration: Planned for future schema extension

BIBFRAME (Bibliographic Resources):

Use for: Library catalogs, bibliographic metadata
Key classes: bf:Organization, bf:Work, bf:Instance
Integration: For library-specific extensions

Reference Documentation: See /docs/ONTOLOGY_EXTENSIONS.md for comprehensive integration patterns, RDF serialization examples, and extension workflows.

Institution Type Taxonomy

The project uses a 19-type GLAMORCUBESFIXPHDNT taxonomy (expanded November 2025) with single-letter codes for GHCID identifier generation:

Type	Code	Description	Example Use Cases
GALLERY	G	Art gallery or exhibition space	Commercial galleries, kunsthallen
LIBRARY	L	Library (public, academic, specialized)	National libraries, university libraries
ARCHIVE	A	Archive (government, corporate, personal)	National archives, city archives
MUSEUM	M	Museum (art, history, science, etc.)	Rijksmuseum, natural history museums
OFFICIAL_INSTITUTION	O	Government heritage agencies	Provincial archives, heritage platforms
RESEARCH_CENTER	R	Research institutes and documentation centers	Knowledge centers, research libraries
CORPORATION	C	Corporate heritage collections	Company archives, corporate museums
UNKNOWN	U	Institution type cannot be determined	Ambiguous or unclassifiable organizations
BOTANICAL_ZOO	B	Botanical gardens and zoological parks	Arboreta, botanical gardens, zoos
EDUCATION_PROVIDER	E	Educational institutions with collections	Schools, training centers with heritage materials, universities
COLLECTING_SOCIETY	S	Societies collecting specialized materials	Numismatic societies, heritage societies (heemkundige kring)
FEATURES	F	Physical landscape features with heritage significance	Monuments, sculptures, statues, memorials, landmarks, cemeteries
INTANGIBLE_HERITAGE_GROUP	I	Organizations preserving intangible heritage	Traditional performance groups, oral history societies, folklore organizations
MIXED	X	Multiple types (uses X code)	Combined museum/archive facilities
PERSONAL_COLLECTION	P	Private personal collections	Individual collectors
HOLY_SITES	H	Religious heritage sites and institutions	Churches, temples, mosques, synagogues with collections
DIGITAL_PLATFORM	D	Digital heritage platforms and repositories	Online archives, digital libraries, virtual museums
NGO	N	Non-governmental heritage organizations	Heritage advocacy groups, preservation societies
TASTE_SMELL	T	Culinary and olfactory heritage institutions	Historic restaurants, parfumeries, distilleries preserving traditional recipes and formulations

Notes:

MIXED institutions use "X" as the GHCID code and document all actual types in metadata
HOLY_SITES includes religious institutions managing cultural heritage collections (archives, libraries, artifacts)
FEATURES includes physical monuments and landscape features with heritage value (not institutions maintaining collections)
COLLECTING_SOCIETY includes historical societies (historische vereniging), philatelic societies, numismatic clubs, ephemera collectors
OFFICIAL_INSTITUTION includes aggregation platforms, provincial heritage services, and government heritage agencies
INTANGIBLE_HERITAGE_GROUP covers organizations preserving UNESCO-recognized intangible cultural heritage
DIGITAL_PLATFORM includes born-digital heritage platforms and digitization aggregators
NGO includes non-profit heritage organizations that don't fit other categories
TASTE_SMELL includes establishments actively preserving culinary traditions, historic recipes, perfume formulations, and sensory heritage
When institution type is unknown, records default to UNKNOWN pending verification

Mnemonic: GLAMORCUBESFIXPHDNT - Galleries, Libraries, Archives, Museums, Official institutions, Research centers, Corporations, Unknown, Botanical gardens/zoos, Education providers, Societies, Features, Intangible heritage groups, miXed, Personal collections, Holy sites, Digital platforms, NGOs, Taste/smell heritage

Note on order: The mnemonic GLAMORCUBESFIXPHDNT represents the alphabetical ordering by code: G-L-A-M-O-R-C-U-B-E-S-F-I-X-P-H-D-N-T

Note: Universities are classified under E (EDUCATION_PROVIDER), not U. The U-class is reserved for institutions where the type cannot be determined during data extraction.

Data Sources

Primary Sources

Conversation JSON files (/Users/kempersc/Documents/claude/glam/*.json)
- 139 conversation files covering global GLAM research
- Countries include: Brazil, Vietnam, Chile, Japan, Mexico, Norway, Thailand, Taiwan, Belgium, Azerbaijan, Estonia, Namibia, Argentina, Tunisia, Ghana, Iran, Russia, Uzbekistan, Armenia, Georgia, Croatia, Greece, Nigeria, Somalia, Yemen, Oman, South Korea, Malaysia, Colombia, Switzerland, Moldova, Romania, Albania, Bosnia, Pakistan, Suriname, Nicaragua, Congo, Denmark, Austria, Australia, Myanmar, Cambodia, Sri Lanka, Tajikistan, Turkmenistan, Philippines, Latvia, Palestine, Limburg (NL), Gelderland (NL), Drenthe (NL), Groningen (NL), Slovakia, Kenya, Paraguay, Honduras, Mozambique, Eritrea, Sudan, Rwanda, Kiribati, Jamaica, Indonesia, Italy, Zimbabwe, East Timor, UAE, Kuwait, Lebanon, Syria, Maldives, Benin
- Also 14 ontology research conversations
Dutch ISIL Registry (data/ISIL-codes_2025-08-01.csv)
- ~300 Dutch heritage institutions
- Fields: Volgnr, Plaats, Instelling, ISIL code, Toegekend op, Opmerking
- Authoritative source (Tier 1)
Dutch Organizations CSV (data/voorbeeld_lijst_organisaties_en_diensten-totaallijst_nederland.csv)
- Comprehensive Dutch heritage organizations
- 40+ metadata columns including: name, address, ISIL code, organization type, partnerships, systems used, metadata standards
- Rich integration data (Museum register, Rijkscollectie, Collectie Nederland, Archieven.nl, etc.)
- Authoritative source (Tier 1)

Implementation Status (Updated Nov 2025)

Both Dutch datasets have been successfully parsed and cross-linked:

ISIL Registry ✅:

364 institutions parsed (2 invalid codes rejected)
203 cities covered
Parser: src/glam_extractor/parsers/isil_registry.py
Tests: 10/10 passing (84% coverage)

Dutch Organizations ✅:

1,351 institutions parsed
475 cities covered
1,119 organizations with digital platforms
Parser: src/glam_extractor/parsers/dutch_orgs.py
Tests: 18/18 passing (98% coverage)

Cross-linking Results 🔗:

340 institutions matched by ISIL code (92.1% overlap)
198 records enriched with platform data
127 name conflicts detected (require manual review)
1,004 organizations without ISIL codes (candidates for assignment)

Analysis Scripts:

compare_dutch_datasets.py - Dataset comparison
crosslink_dutch_datasets.py - TIER_1 data merging demo
test_real_dutch_orgs.py - Real data validation

See PROGRESS.md for detailed statistics and findings.

Conversation JSON Structure

Each conversation JSON file has the following structure:

{
  "uuid": "conversation-uuid",
  "name": "Conversation name (often includes country/region)",
  "summary": "Optional summary",
  "created_at": "ISO 8601 timestamp",
  "updated_at": "ISO 8601 timestamp",
  "chat_messages": [
    {
      "uuid": "message-uuid",
      "text": "User or assistant message text",
      "sender": "human" | "assistant",
      "content": [
        {
          "type": "text" | "tool_use" | "tool_result",
          "text": "Message content (may contain markdown, lists, etc.)",
          ...
        }
      ]
    }
  ]
}

NLP Extraction Tasks

All extraction tasks map to the modular LinkML schema v0.2.0. See Schema Reference section above for module details.

Task 1: Entity Recognition - Institution Names

Objective: Extract heritage institution names from conversation text.

Schema Mapping: Populates HeritageCustodian class from schemas/core.yaml

Patterns to Look For:

Organization names (proper nouns)
Museum names (often contain "Museum", "Museu", "Museo", "Muzeum", etc.)
Library names (contain "Library", "Biblioteca", "Bibliothek", "Bibliotheek", etc.)
Archive names (contain "Archive", "Archivo", "Archiv", "Archief", etc.)
Gallery names
Cultural centers
Holy sites with collections (churches, temples, mosques, synagogues, monasteries, abbeys, cathedrals managing heritage materials)

Contextual Indicators:

Lists of institutions
Descriptions like "The X is a museum in Y"
URLs containing institution names
Mentions of collections, exhibitions, or holdings

Example Extraction:

Input: "The Biblioteca Nacional do Brasil in Rio de Janeiro holds over 9 million items..."

Output:
- name: "Biblioteca Nacional do Brasil"  # HeritageCustodian.name
- institution_type: LIBRARY  # InstitutionTypeEnum from schemas/enums.yaml
- city: "Rio de Janeiro"  # Location.city from schemas/core.yaml
- confidence_score: 0.95  # Provenance.confidence_score from schemas/provenance.yaml

Task 2: Location Extraction

Objective: Extract geographic information associated with institutions.

Schema Mapping: Populates Location class from schemas/core.yaml

Extract:

City names
Street addresses (when mentioned)
Postal codes
Provinces/states/regions
Country (can often be inferred from conversation title)

Geocoding:

Use Nominatim API to geocode addresses to lat/lon
Link to GeoNames IDs when possible
Handle multilingual place names

Example:

Input: "Nationaal Onderduikmuseum, Aalten"

Output:
- city: "Aalten"  # Location.city
- country: "NL"  # Location.country (ISO 3166-1 alpha-2)
- geonames_id: "2759899" (lookup via API)  # Location.geonames_id
- latitude: 51.9167 (from geocoding)
- longitude: 6.5833

Task 3: Identifier Extraction

Objective: Extract external identifiers mentioned in conversations.

Schema Mapping: Populates Identifier class from schemas/core.yaml

Identifier Types:

ISIL codes (format: NL-XXXXX, US-XXXXX, etc.)
Wikidata IDs (format: Q12345)
VIAF IDs (format: numeric)
URLs to institutional websites
KvK numbers (Dutch: 8-digit format)

Patterns:

ISIL: [A-Z]{2}-[A-Za-z0-9]+
Wikidata: Q[0-9]+
VIAF: viaf.org/viaf/[0-9]+
KvK: [0-9]{8}

Example:

Input: "ISIL code NL-AsdAM for Amsterdam Museum"

Output:
- identifier_scheme: "ISIL"  # Identifier.identifier_scheme
- identifier_value: "NL-AsdAM"  # Identifier.identifier_value
- institution_name: "Amsterdam Museum"  # HeritageCustodian.name (for linking)

Task 4: Relationship Extraction

Objective: Extract relationships between institutions.

Schema Mapping: Maps to ChangeEvent class from schemas/provenance.yaml (for mergers, splits) and future relationship modeling

Relationship Types:

Parent-child (e.g., "X is part of Y")
Partnerships (e.g., "X collaborates with Y")
Network memberships (e.g., "X is a member of Z consortium")
Merged organizations (e.g., "X merged with Y") → ChangeTypeEnum.MERGER

Indicators:

"part of", "branch of", "division of"
"in partnership with", "collaborates with"
"member of", "belongs to"
"merged with", "absorbed by" → Use ChangeEvent from schemas/provenance.yaml

Task 5: Collection Metadata Extraction

Objective: Extract information about collections held by institutions.

Schema Mapping: Populates Collection class from schemas/collections.yaml

Extract:

Collection names → Collection.collection_name
Collection types (archival, bibliographic, museum objects)
Subject areas → Collection.subject_areas
Time periods covered → Collection.temporal_coverage
Item counts (when mentioned) → Collection.extent
Access information → Collection.access_rights

Example:

Input: "The archive holds 15,000 documents from the 18th-19th centuries..."

Output:
- collection_type: "archival"  # Collection metadata
- item_count: 15000  # Collection.extent
- time_period_start: "1700-01-01"  # Collection.temporal_coverage
- time_period_end: "1899-12-31"

Task 6: Digital Platform Identification

Objective: Identify digital platforms and systems used by institutions.

Schema Mapping: Populates DigitalPlatform class from schemas/core.yaml

Platform Types:

Collection management systems (Atlantis, MAIS, CollectiveAccess, etc.)
Digital repositories (DSpace, EPrints, Fedora)
Discovery portals
SPARQL endpoints
APIs

Extract:

Platform name → DigitalPlatform.platform_name
Platform URL → DigitalPlatform.platform_url
Metadata standards used → DigitalPlatform.metadata_standards
Integration with aggregators (Europeana, DPLA, etc.)

Task 7: Metadata Standards Detection

Objective: Identify which metadata standards institutions use.

Schema Mapping: Stores in DigitalPlatform.metadata_standards (list of strings)

Standards to Detect:

Dublin Core
MARC21
EAD (Encoded Archival Description)
BIBFRAME
LIDO
CIDOC-CRM
Schema.org
RiC-O (Records in Contexts)
MODS, PREMIS, SPECTRUM, DACS

Indicators:

Explicit mentions: "uses Dublin Core", "MARC21 records"
Implicit: technical discussions about cataloging practices

Task 8: Organizational Change Event Extraction (NEW - v0.2.0)

Objective: Extract significant organizational change events from conversation history.

Schema Mapping: Populates ChangeEvent class from schemas/provenance.yaml

Change Types to Detect (from ChangeTypeEnum in schemas/enums.yaml):

FOUNDING: "established", "founded", "created", "opened"
CLOSURE: "closed", "dissolved", "ceased operations", "shut down"
MERGER: "merged with", "combined with", "joined with", "absorbed"
SPLIT: "split into", "divided into", "separated from", "spun off"
ACQUISITION: "acquired", "took over", "purchased"
RELOCATION: "moved to", "relocated to", "transferred to"
NAME_CHANGE: "renamed to", "formerly known as", "changed name to"
TYPE_CHANGE: "became a museum", "converted to archive", "now operates as"
STATUS_CHANGE: "reopened", "temporarily closed", "suspended operations"
RESTRUCTURING: "reorganized", "restructured", "reformed"
LEGAL_CHANGE: "incorporated as", "became a foundation", "legal status changed"

Extract for Each Event:

change_history:  # HeritageCustodian.change_history (list of ChangeEvent)
  - event_id: "https://w3id.org/heritage/custodian/event/unique-id"  # ChangeEvent.event_id
    change_type: MERGER  # ChangeEvent.change_type (ChangeTypeEnum from schemas/enums.yaml)
    event_date: "2001-01-01"  # ChangeEvent.event_date
    event_description: >-  # ChangeEvent.event_description
      Merger of Institution A and Institution B to form new organization C.
      Detailed description from conversation.
    affected_organization: null  # ChangeEvent.affected_organization (optional)
    resulting_organization: null  # ChangeEvent.resulting_organization (optional)
    related_organizations: []  # ChangeEvent.related_organizations (optional)
    source_documentation: "https://..."  # ChangeEvent.source_documentation (optional)

Temporal Context Indicators:

"In 2001, the museum merged with..."
"After the renovation in 1985..."
"Following the name change in 1968..."
"The archive was relocated from X to Y in 1923"

PROV-O Integration:

Map to prov:Activity in RDF serialization
Link with prov:wasInfluencedBy from HeritageCustodian
Use prov:atTime for event timestamps
Track prov:entity (affected) and prov:generated (resulting) organizations

Example Extraction:

Input: "The Noord-Hollands Archief was formed in 2001 through a merger of 
        Gemeentearchief Haarlem (founded 1910) and Rijksarchief in Noord-Holland 
        (founded 1802). The merger created a unified regional archive serving both 
        the city and province."

Output:
- event_id: "https://w3id.org/heritage/custodian/event/nha-merger-2001"
- change_type: MERGER  # ChangeTypeEnum.MERGER
- event_date: "2001-01-01"
- event_description: "Merger of Gemeentearchief Haarlem (municipal archive, founded 
                      1910) and Rijksarchief in Noord-Holland (state archive, founded 
                      1802) to form Noord-Hollands Archief."
- confidence_score: 0.95  # From Provenance metadata

GHCID Impact:

When institutions merge, relocate, or change names, GHCID may change
Track old GHCID in ghcid_history with valid_to timestamp matching event date → GHCIDHistoryEntry from schemas/provenance.yaml
Create new GHCIDHistoryEntry with valid_from matching event date
Link change event to GHCID change via temporal correlation

Indicators:

Task 9: Holy Sites Heritage Collection Identification

Objective: Identify religious sites that function as heritage custodians by maintaining cultural collections.

Schema Mapping: Populates HeritageCustodian class with institution_type: HOLY_SITES

When to Classify as HOLY_SITES:

Religious institutions qualify as HOLY_SITES heritage custodians when they manage:

Archival collections: Historical documents, parish registers, ecclesiastical records
Library collections: Rare manuscripts, theological texts, historical books
Museum collections: Religious artifacts, liturgical objects, art collections
Cultural heritage: Historical buildings with guided tours, preservation programs

Patterns to Look For:

Church archives (parish records, baptismal registers, historical documents)
Monastery libraries (manuscript collections, rare books)
Cathedral treasuries (liturgical objects, religious art)
Temple museums (Buddhist artifacts, historical collections)
Mosque libraries (Islamic manuscripts, Quranic texts)
Synagogue archives (Jewish community records, Torah scrolls)
Abbey collections (medieval manuscripts, historical artifacts)

Keywords and Indicators:

"church archive", "parish records", "ecclesiastical archive"
"monastery library", "monastic collection", "scriptorium"
"cathedral treasury", "cathedral museum"
"temple library", "temple collection"
"mosque library", "Islamic manuscript collection"
"synagogue archive", "Jewish heritage collection"
"religious heritage site", "pilgrimage site with museum"

NOT Holy Sites (use other types):

Secular museums about religion (use MUSEUM)
Academic religious studies centers (use RESEARCH_CENTER or UNIVERSITY)
Government archives of church records (use ARCHIVE)
Religious organizations without heritage collections (not heritage custodians)

Example Extraction:

Input: "The Vatican Apostolic Archive holds over 85 km of shelving with 
        documents dating back to the 8th century, including papal bulls, 
        correspondence, and medieval manuscripts."

Output:
- name: Vatican Apostolic Archive
  institution_type: HOLY_SITES  # Religious institution managing heritage collections
  description: >-
    The Vatican Apostolic Archive (formerly Vatican Secret Archives) is 
    the central repository for papal and Vatican documents, holding over 
    35,000 volumes of historical records spanning 12 centuries.    
  locations:
    - city: Vatican City
      country: VA
  collections:
    - collection_name: Papal Documents
      collection_type: archival
      temporal_coverage: "0800-01-01/2024-12-31"
      extent: "85 kilometers of shelving, 35,000+ volumes"
  provenance:
    data_source: CONVERSATION_NLP
    confidence_score: 0.95

Schema.org Mapping:

HOLY_SITES maps to schema:PlaceOfWorship in RDF serialization
Can also use schema:ArchiveOrganization or schema:Library for collection-specific context
Use multiple type assertions when appropriate

Cross-Cultural Considerations:

Christianity: churches, cathedrals, monasteries, abbeys, convents
Islam: mosques, madrasas (with historical libraries)
Judaism: synagogues, yeshivas (with archival collections)
Buddhism: temples, monasteries, pagodas (with artifact collections)
Hinduism: temples (with historical collections)
Sikhism: gurdwaras (with historical manuscripts)
Other faiths: shrines, pilgrimage sites with documented heritage collections

Data Quality and Provenance

Provenance Tracking

Every extracted record MUST include:

provenance:
  data_source: CONVERSATION_NLP
  data_tier: TIER_4_INFERRED
  extraction_date: "2025-11-05T..."
  extraction_method: "Subagent NER + pattern matching"
  confidence_score: 0.85
  conversation_id: "conversation-uuid"
  source_url: null
  verified_date: null
  verified_by: null

Confidence Scoring

Assign confidence scores (0.0-1.0) based on:

0.9-1.0: Explicit, unambiguous mentions with context
0.7-0.9: Clear mentions with some ambiguity
0.5-0.7: Inferred from context, may need verification
0.3-0.5: Low confidence, likely needs verification
0.0-0.3: Very uncertain, flag for manual review

Data Tier Assignment

TIER_1_AUTHORITATIVE: CSV registries (ISIL, Dutch orgs)
TIER_2_VERIFIED: Data from institutional websites (crawl4ai)
TIER_3_CROWD_SOURCED: Wikidata, OpenStreetMap
TIER_4_INFERRED: NLP-extracted from conversations

Integration with CSV Data

Cross-linking Strategy

ISIL Code Matching (primary)
- If conversation mentions ISIL code, link to CSV record
- High confidence match
Name Matching (secondary)
- Normalize names (lowercase, remove punctuation, handle abbreviations)
- Fuzzy matching with threshold > 0.85
- Check for alternative names
Location + Type Matching (tertiary)
- Match by city + institution type
- Lower confidence, requires manual verification

Conflict Resolution

When conversation data conflicts with CSV data:

CSV data takes precedence (higher tier)
Mark conversation data with verified: false
Note conflict in provenance metadata
Create separate record if institutions are genuinely different

NLP Models and Tools

Recommended Approach: Agent-Based NER

IMPORTANT: Instead of directly using spaCy or other NER libraries in the main codebase, use coding subagents via the Task tool to conduct Named Entity Recognition and text extraction.

Why Subagents:

Keeps the main codebase clean and maintainable
Allows flexible experimentation with different NER approaches
Subagents can choose the best tool for each specific extraction task
Better separation of concerns: extraction logic vs. data pipeline

How to Use Subagents for NER:

Use the Task tool with subagent_type="general" for NER tasks
Provide clear prompts describing what entities to extract
Subagent will autonomously choose and apply appropriate NER tools (spaCy, transformers, regex, etc.)
Subagent returns structured extraction results
Main code validates and processes the results

CRITICAL: Creating LinkML Instance Files

Agent Capabilities Go Beyond Traditional NER

IMPORTANT: AI extraction agents are NOT limited to simple Named Entity Recognition. Unlike traditional NER tools that only identify entity boundaries and types, AI agents have comprehensive understanding and can:

Extract Complete Records: Capture ALL relevant information for each institution in one pass
Infer Missing Data: Use context to fill in fields that aren't explicitly stated
Cross-Reference Within Documents: Link related entities (locations, identifiers, events) automatically
Maintain Consistency: Ensure all extracted data conforms to the LinkML schema
Generate Rich Metadata: Create complete provenance tracking and confidence scores

Mandatory: Create Complete LinkML Instance Files

When extracting data from conversations or other sources, agents MUST:

✅ DO THIS: Create complete LinkML-compliant YAML instance files with ALL available information

# Example: data/instances/brazil_museums_001.yaml
---
# From schemas/core.yaml - HeritageCustodian class

- id: https://w3id.org/heritage/custodian/br/bnb-001
  name: Biblioteca Nacional do Brasil
  institution_type: LIBRARY  # From schemas/enums.yaml
  alternative_names:
    - National Library of Brazil
    - BNB
  description: >-
    The National Library of Brazil, located in Rio de Janeiro, is the largest 
    library in Latin America with over 9 million items. Founded in 1810 by 
    King João VI of Portugal. Collections include rare manuscripts, maps, 
    photographs, and Brazilian historical documents.    
  
  locations:  # From schemas/core.yaml - Location class
    - city: Rio de Janeiro
      street_address: Avenida Rio Branco, 219
      postal_code: "20040-008"
      region: Rio de Janeiro
      country: BR
      # Note: lat/lon can be geocoded later if not in text
  
  identifiers:  # From schemas/core.yaml - Identifier class
    - identifier_scheme: ISIL
      identifier_value: BR-RjBN
      identifier_url: https://isil.org/BR-RjBN
    - identifier_scheme: VIAF
      identifier_value: "123556639"
      identifier_url: https://viaf.org/viaf/123556639
    - identifier_scheme: Wikidata
      identifier_value: Q1526131
      identifier_url: https://www.wikidata.org/wiki/Q1526131
    - identifier_scheme: Website
      identifier_value: https://www.bn.gov.br
      identifier_url: https://www.bn.gov.br
  
  digital_platforms:  # From schemas/core.yaml - DigitalPlatform class
    - platform_name: Digital Library of the National Library of Brazil
      platform_url: https://bndigital.bn.gov.br
      platform_type: DISCOVERY_PORTAL
      metadata_standards:
        - Dublin Core
        - MARC21
  
  collections:  # From schemas/collections.yaml - Collection class
    - collection_name: Brazilian Historical Documents
      collection_type: archival
      subject_areas:
        - Brazilian History
        - Colonial Period
        - Imperial Brazil
      temporal_coverage: "1500-01-01/1889-11-15"
      extent: "Approximately 2.5 million documents"
  
  change_history:  # From schemas/provenance.yaml - ChangeEvent class
    - event_id: https://w3id.org/heritage/custodian/event/bnb-founding-1810
      change_type: FOUNDING
      event_date: "1810-01-01"
      event_description: >-
        Founded by King João VI of Portugal as the Royal Library 
        (Biblioteca Real) when the Portuguese court relocated to Brazil.        
      source_documentation: https://www.bn.gov.br/sobre-bn/historia
  
  provenance:  # From schemas/provenance.yaml - Provenance class
    data_source: CONVERSATION_NLP
    data_tier: TIER_4_INFERRED
    extraction_date: "2025-11-05T14:30:00Z"
    extraction_method: "AI agent comprehensive extraction from Brazilian GLAM conversation"
    confidence_score: 0.92
    conversation_id: "2025-09-22T14-40-15-0102c00a-4c0a-4488-bdca-5dd9fb94c9c5"
    notes: >-
      Extracted from conversation about Brazilian GLAM institutions. 
      Historical founding information cross-referenced from institutional website.

❌ DO NOT DO THIS: Return minimal JSON with only name and type

// BAD - This is insufficient!
{
  "name": "Biblioteca Nacional do Brasil",
  "institution_type": "LIBRARY"
}

Extraction Workflow for Agents

When processing a conversation or document:

Read Entire Document First: Don't extract piecemeal - understand the full context
Identify ALL Entities: Find every institution, location, identifier, event mentioned
Gather Complete Information: For each institution, extract:
- Basic metadata (name, type, description)
- All locations mentioned (even if just city/country)
- All identifiers (ISIL, Wikidata, VIAF, URLs)
- Digital platforms and systems
- Collection information
- Historical events (founding, mergers, relocations)
- Relationships to other institutions
Create LinkML YAML: Write a complete instance file with ALL extracted data
Add Provenance: Always include extraction metadata with confidence scores
Validate: Ensure output conforms to schema (use linkml-validate if available)

Example Agent Prompt for Comprehensive Extraction

Extract ALL heritage institutions from the following conversation about Brazilian GLAM institutions.

For EACH institution found, create a COMPLETE LinkML-compliant record including:
- Institution name, type, and description
- ALL locations mentioned (cities, addresses, regions)
- ALL identifiers (ISIL codes, Wikidata IDs, VIAF IDs, URLs)
- Digital platforms, systems, or websites
- Collection information (types, subjects, time periods, extent)
- Historical events (founding dates, mergers, relocations, name changes)
- Relationships to other organizations

Output: YAML file conforming to schemas/core.yaml, schemas/enums.yaml, 
schemas/provenance.yaml, and schemas/collections.yaml

Use your understanding to:
- Infer missing fields from context (e.g., country from city names)
- Consolidate information scattered across multiple conversation turns
- Create rich descriptions summarizing key facts
- Assign appropriate confidence scores based on explicitness of mentions

Remember: You are NOT a simple NER tool. Use your full comprehension abilities 
to create the most complete, accurate, and useful records possible.

Multiple Institutions Per File

When a conversation discusses many institutions, create ONE YAML file with a list:

---
# data/instances/netherlands_limburg_museums.yaml

- id: https://w3id.org/heritage/custodian/nl/bonnefantenmuseum
  name: Bonnefantenmuseum
  institution_type: MUSEUM
  # ... complete record ...

- id: https://w3id.org/heritage/custodian/nl/thermenmuseum
  name: Thermenmuseum
  institution_type: MUSEUM
  # ... complete record ...

- id: https://w3id.org/heritage/custodian/nl/limburgs-museum
  name: Limburgs Museum
  institution_type: MUSEUM
  # ... complete record ...

Field Completion Strategies

Even when information is incomplete, do your best:

No explicit institution type? Infer from context ("national library" → LIBRARY)
Only city mentioned? That's fine - add locations: [{city: "Amsterdam", country: "NL"}]
No ISIL code? Check if you can infer the format (NL-CityCode) or leave it out
No description? Create one from available facts
Uncertain data? Lower the confidence score but still include it

Validation and Quality Control

After creating instance files:

Schema Validation: If possible, run linkml-validate -s schemas/heritage_custodian.yaml data/instances/your_file.yaml
Completeness Check: Ensure every institution has at minimum:
- id (generate from country + institution name slug)
- name
- institution_type
- provenance (with data_source, extraction_date, confidence_score)
Consistency Check: Same institution mentioned multiple times? Merge into one record
Quality Flags: If confidence < 0.5, add note in provenance.notes explaining uncertainty

Extraction Stack (for Subagents)

When subagents perform extraction, they may use:

Pattern matching for identifiers (primary approach)
- Regex for ISIL, VIAF, Wikidata IDs
- URL extraction and normalization
- High precision, no dependencies
NER libraries (via subagents only)
- spaCy: en_core_web_trf, nl_core_news_lg, xx_ent_wiki_sm
- Transformers for classification
- Used by subagents, not directly in main code
Fuzzy matching for deduplication
- rapidfuzz library
- Levenshtein distance for name matching

Processing Pipeline

Conversation JSON
    ↓
Parse & Extract Text
    ↓
[SUBAGENT] NER Extraction
  - Subagent uses spaCy/transformers/patterns
  - Returns structured entities
    ↓
Pattern Matching (identifiers, URLs)
    ↓
Classification (institution type, standards)
    ↓
Geocoding (locations)
    ↓
Cross-link with CSV (ISIL/name matching)
    ↓
LinkML Validation
    ↓
Export (RDF, JSON-LD, CSV, Parquet)

Agent Interaction Patterns

When Asked to Extract Data from Conversations

Start Small: Begin with 1-2 conversation files to test extraction logic
Show Examples: Display extracted entities with confidence scores
Ask for Validation: Show uncertain extractions for user confirmation
Iterate: Refine patterns based on feedback
Batch Process: Once patterns are validated, process all 139 files

When Asked to Design NLP Components

Reference Schema: Always refer to the modular schema v0.2.1:
- Core classes: schemas/core.yaml (HeritageCustodian, Location, Identifier, etc.)
- Enumerations: schemas/enums.yaml (InstitutionTypeEnum, ChangeTypeEnum, etc.)
- Provenance: schemas/provenance.yaml (Provenance, ChangeEvent, etc.)
- See schema overview in the "Schema Reference (v0.2.1)" section above
Consult Base Ontologies: BEFORE designing extraction logic, review relevant ontologies:
- Dutch institutions: Study TOOI ontology (/data/ontology/tooiont.ttl)
- EU/global institutions: Study CPOV ontology (/data/ontology/core-public-organisation-ap.ttl)
- All institutions: Reference Schema.org patterns (/data/ontology/schemaorg.owl)
- See "Base Ontologies for Global GLAM Data" section above for decision tree
Use Design Patterns: Follow patterns in docs/plan/global_glam/05-design-patterns.md
Track Provenance: Every extraction must include provenance metadata (from schemas/provenance.yaml)
Handle Multilingual: Conversations cover 60+ countries, expect multilingual content
Error Handling: Use Result pattern, never fail silently

When Asked to Validate Data

LinkML Validation: Use linkml-validate to check schema compliance
Cross-reference: Compare with CSV data when applicable
Check Identifiers: Validate ISIL format, check Wikidata exists
Geographic Verification: Geocode addresses, verify country codes
Duplicate Detection: Use fuzzy matching to find potential duplicates

Example Agent Workflows

Workflow 1: Extract Brazilian Institutions

# User request
"Extract all museum, library, and archive names from the Brazilian GLAM conversation"

# Agent actions
1. Read conversation: 2025-09-22T14-40-15-0102c00a-4c0a-4488-bdca-5dd9fb94c9c5-Brazilian_GLAM_collection_inventories.json
2. Parse chat_messages array
3. **Launch subagent** to extract institutions using NER
   - Subagent analyzes text and extracts ORG entities
   - Filters for heritage-related keywords
   - Classifies institution types
   - Returns structured results
4. Extract locations (cities in Brazil)
5. Geocode using Nominatim
6. Create HeritageCustodian records
7. Add provenance metadata (data_source: CONVERSATION_NLP, extraction_method: "Subagent NER")
8. Validate with LinkML schema
9. Export to JSON-LD
10. Report results with confidence scores

Workflow 2: Cross-link Dutch Institutions

# User request
"Cross-link the Dutch organizations CSV with any Dutch institutions found in conversations"

# Agent actions
1. Load data/voorbeeld_lijst_organisaties_en_diensten-totaallijst_nederland.csv
2. Parse into DutchHeritageCustodian records
3. Extract all NL-* ISIL codes
4. Search all conversation files for mentions of these ISIL codes
5. Fuzzy match organization names
6. For matches:
   - Merge metadata
   - Mark CSV data as TIER_1
   - Mark conversation data as TIER_4
   - Resolve conflicts (CSV wins)
7. For Dutch institutions in conversations NOT in CSV:
   - Create new records
   - Mark as TIER_4
   - Flag for verification
8. Export merged dataset

Workflow 3: Build Global Institution Map

# User request
"Create a geographic distribution map of all extracted institutions"

# Agent actions
1. Process all 139 conversation files
2. **Launch subagent(s)** to extract institution names + locations from each file
3. Geocode all addresses
4. Group by country
5. Count institutions per country
6. Generate GeoJSON for mapping
7. Create visualization (Leaflet, Mapbox, etc.)
8. Export statistics:
   - Institutions per country
   - Institutions per type
   - Geographic coverage
   - Data quality (tier distribution)

Multi-language Considerations

Language Detection

Detect language of conversation content
Subagents will choose appropriate NER models per language
Multilingual support handled by subagents

Common Languages in Dataset

English (international institutions)
Dutch (Netherlands institutions)
Portuguese (Brazil)
Spanish (Latin America, Spain)
Vietnamese, Japanese, Thai, Korean, Arabic, Russian, etc.

Translation Strategy

DO NOT translate institution names (preserve original)
Optionally translate descriptions for searchability
Store language tags with text fields
Use multilingual identifiers (Wikidata) for linking

Output Formats

Primary Output: JSON-LD

Linked Data format for semantic web integration:

{
  "@context": "https://w3id.org/heritage/custodian/context.jsonld",
  "@type": "HeritageCustodian",
  "@id": "https://example.org/institution/123",
  "name": "Amsterdam Museum",
  "institution_type": "MUSEUM",
  ...
}

Secondary Outputs

RDF/Turtle: For SPARQL querying
CSV: For spreadsheet analysis
Parquet: For data warehousing
SQLite: For local querying

Testing and Validation

Unit Tests

Test extraction functions with known inputs:

def test_extract_isil_codes():
    text = "The ISIL code NL-AsdAM identifies Amsterdam Museum"
    codes = extract_isil_codes(text)
    assert codes == [{"scheme": "ISIL", "value": "NL-AsdAM"}]

Integration Tests

Test full pipeline with sample conversations:

def test_brazilian_museum_extraction():
    conversation = load_json("Brazilian_GLAM_collection_inventories.json")
    records = extract_heritage_custodians(conversation)
    assert len(records) > 0
    assert all(r.provenance.data_source == "CONVERSATION_NLP" for r in records)

Validation Tests

Ensure LinkML schema compliance:

def test_linkml_validation():
    record = create_heritage_custodian(...)
    validator = SchemaValidator(schema="heritage_custodian.yaml")
    result = validator.validate(record)
    assert result.is_valid

Performance Optimization

Batch Processing

Process conversations in parallel (multiprocessing)
Cache geocoding results (15-minute TTL)
Deduplicate entity extraction

Incremental Updates

Track last processed timestamp
Only process new/updated conversations
Maintain state in SQLite database

Resource Management

Limit concurrent API calls (Nominatim: 1 req/sec)
Use connection pooling for HTTP requests
Stream large JSON files instead of loading into memory

Error Handling

Common Errors and Solutions

JSON Parsing Errors
- Malformed JSON files
- Solution: Validate JSON schema, report file path
NER Model Errors
- Missing spaCy model
- Solution: Provide installation instructions, download automatically
Geocoding Failures
- Unknown location, rate limit exceeded
- Solution: Cache results, implement backoff, mark as unverified
LinkML Validation Failures
- Required field missing, invalid enum value
- Solution: Log validation errors, provide field mapping
Encoding Issues
- Non-UTF-8 characters
- Solution: Use UTF-8 everywhere, handle decode errors gracefully

Schema Quirks and Implementation Notes

IMPORTANT: These are critical implementation details discovered during development. Read carefully to avoid bugs.

Provenance Model Quirks

The Provenance model does NOT have a notes field:

# ❌ WRONG - Provenance has no 'notes' field
provenance = Provenance(
    data_source=DataSource.CSV_REGISTRY,
    notes="Some observation"  # This will fail!
)

# ✅ CORRECT - Use HeritageCustodian.description instead
custodian = HeritageCustodian(
    name="Museum Name",
    description="Notes and remarks go here",  # Put notes here
    provenance=Provenance(...)
)

Field Naming Conventions

Always use the correct field names (check the schema when in doubt):

# ❌ WRONG
custodian.institution_types  # Plural, list
custodian.location           # Singular

# ✅ CORRECT
custodian.institution_type   # Singular, single enum value
custodian.locations          # Plural, always a list (even with one item)

Pydantic v1 Enum Behavior

This project uses Pydantic v1. Enum fields are already strings, not enum objects:

# ❌ WRONG - Don't use .value accessor
print(custodian.institution_type.value)  # AttributeError!

# ✅ CORRECT - Enum fields are already strings
print(custodian.institution_type)  # "MUSEUM", "ARCHIVE", etc.

# Same for platform types
platform.platform_type  # Already a string, not an enum object

Required vs. Optional Fields

Many fields are optional but have validation rules. Always check for None:

# Optional fields that may be None
custodian.locations          # Optional[List[Location]]
custodian.identifiers        # Optional[List[Identifier]]
custodian.digital_platforms  # Optional[List[DigitalPlatform]]
custodian.description        # Optional[str]

# Always check before iterating
if custodian.locations:
    for location in custodian.locations:
        print(location.city)

CSV Parsing Best Practices

Handle UTF-8 BOM: Use encoding='utf-8-sig' when reading CSVs
Normalize headers: Strip whitespace, handle multiline headers
Warn on errors: Skip invalid rows but log warnings
Preserve originals: Store raw CSV data in intermediate models before conversion

Example:

with open(csv_path, 'r', encoding='utf-8-sig') as f:
    reader = csv.DictReader(f)
    for row in reader:
        try:
            record = parse_row(row)
        except ValidationError as e:
            print(f"Warning: Skipping row {row}: {e}")
            continue

Date Handling

Dates may be in various formats or empty:

# Handle empty dates
date_str = row.get('toegekend_op', '').strip()
assigned_date = datetime.fromisoformat(date_str) if date_str else None

# Provenance extraction_date is required (use current time)
from datetime import datetime, timezone
extraction_date = datetime.now(timezone.utc)

Testing Strategies

Unit tests: Test model validation with known inputs
Integration tests: Test full file parsing with fixtures
Edge case tests: Empty files, malformed rows, minimal data
Real data tests: Always validate with actual CSV files

Fixture scope matters:

# ❌ WRONG - Class-scoped fixture not available to other classes
class TestFoo:
    @pytest.fixture
    def sample_file(self):
        ...

# ✅ CORRECT - Module-scoped fixture available to all test classes
@pytest.fixture
def sample_file():  # At module level, not in a class
    ...

Next Steps for Agents

When continuing this project, agents should:

Implement Parser Module (src/glam_extractor/parsers/) ✅ COMPLETE
- ✅ ISIL registry parser (10 tests, 84% coverage)
- ✅ Dutch organizations parser (18 tests, 98% coverage)
- ⏳ Conversation JSON parser (next priority)
Implement Extractor Module (src/glam_extractor/extractors/)
- spaCy NER integration
- Pattern-based identifier extraction
- Institution type classifier
- Relationship extractor
Implement Geocoder Module (src/glam_extractor/geocoding/)
- Nominatim client with caching
- GeoNames integration
- Coordinate validation
Implement Validator Module (src/glam_extractor/validators/)
- LinkML schema validator
- Cross-reference validator (CSV vs. conversation)
- Duplicate detector
Implement Exporter Module (src/glam_extractor/exporters/)
- JSON-LD exporter
- RDF/Turtle exporter
- CSV exporter
- Parquet exporter
- SQLite database builder
Create Test Fixtures (tests/fixtures/)
- Sample conversation JSONs
- Expected extraction outputs
- Validation test cases
Document Agent Prompts (docs/agent-prompts/)
- Reusable prompts for common extraction tasks
- Few-shot examples for LLM-based extraction
- Quality review checklists

Persistent Identifiers (GHCID)

🚨 COLLISION RESOLUTION: NATIVE LANGUAGE NAME SUFFIX 🚨

When multiple institutions generate the same base GHCID, collisions are resolved by appending the full legal name in native language in snake_case format.

Collision Suffix Rules:

✅ Use the institution's full official name in its native language
✅ Convert to snake_case (lowercase, underscores for spaces)
✅ Remove apostrophes, accents, commas, and other punctuation/diacritics
✅ Only add suffix on collision (not by default)
✅ First-added institution keeps base GHCID; later additions get name suffix

Examples:

Base GHCID collision: NL-NH-AMS-M-SM (two museums with "SM" abbreviation)
First institution: NL-NH-AMS-M-SM (Stedelijk Museum, added first - no suffix)
Second institution: NL-NH-AMS-M-SM-science_museum_amsterdam (added later - gets suffix)

Name Normalization:

"Musée d'Orsay" → "musee_dorsay"
"Biblioteca Nacional do Brasil" → "biblioteca_nacional_do_brasil"
"北京故宫博物院" → "beijing_gugong_bowuyuan" (pinyin transliteration)
"Österreichische Nationalbibliothek" → "osterreichische_nationalbibliothek"

Note: The GHCID string (including any name suffix) gets hashed to UUID, so the longer name won't be visible to end users - they see only the UUID.

🚨 SETTLEMENT STANDARDIZATION: GEONAMES IS AUTHORITATIVE 🚨

ALL settlement names in GHCID MUST be derived from GeoNames, not from source data.

The GeoNames geographical database (/data/reference/geonames.db) is the single source of truth for:

Settlement names (cities, towns, villages)
Settlement 3-letter codes
Administrative region codes (admin1 → ISO 3166-2)

Why GeoNames?

Consistency: Same coordinates → same settlement → same GHCID component
Disambiguation: Handles duplicate settlement names across regions
Standardization: Provides ASCII-safe names for identifiers
Persistence: Geographic reality is stable, ensuring GHCID stability

Settlement Resolution Process:

Coordinates Available (Preferred): Use reverse geocoding to find nearest GeoNames settlement
Name Only (Fallback): Look up settlement name in GeoNames with fuzzy matching
Manual (Last Resort): Flag entry with settlement_code: XXX for review

🚨 CRITICAL: GeoNames Feature Code Filtering 🚨

NEVER use neighborhoods or districts (PPLX) for GHCID generation. ONLY use proper settlements (cities, towns, villages).

GeoNames classifies populated places with feature codes. When reverse geocoding coordinates to find a settlement, you MUST filter by feature code to ensure you get a city/town/village, NOT a neighborhood or district.

ALLOWED Feature Codes (use these for GHCID settlements):

Code	Description	Example
PPL	Populated place (city/town/village)	Apeldoorn, Hamont, Lelystad
PPLA	Seat of first-order admin division	Provincial capitals
PPLA2	Seat of second-order admin division	Municipal seats
PPLA3	Seat of third-order admin division	District seats
PPLA4	Seat of fourth-order admin division	Sub-district seats
PPLC	Capital of a political entity	Amsterdam, Brussels
PPLS	Populated places (multiple)	Settlement clusters
PPLG	Seat of government	The Hague (when different from capital)

EXCLUDED Feature Codes (NEVER use for GHCID):

Code	Description	Why Excluded
PPLX	Section of populated place	Neighborhoods, districts, quarters (e.g., "Binnenstad", "Amsterdam Binnenstad")

Example of the Problem:

-- BAD: Query without feature code filter returns neighborhoods
SELECT name, feature_code, population FROM cities 
WHERE country_code='NL' ORDER BY distance LIMIT 1;
-- Result: "Binnenstad" (PPLX, pop 4,900) ❌ WRONG

-- GOOD: Query WITH feature code filter returns proper settlements
SELECT name, feature_code, population FROM cities 
WHERE country_code='NL' 
  AND feature_code IN ('PPL', 'PPLA', 'PPLA2', 'PPLA3', 'PPLA4', 'PPLC', 'PPLS', 'PPLG')
ORDER BY distance LIMIT 1;
-- Result: "Apeldoorn" (PPL, pop 136,670) ✅ CORRECT

Implementation in SQL:

-- Correct reverse geocoding query with feature code filter
SELECT 
    name, ascii_name, admin1_code, admin1_name,
    latitude, longitude, geonames_id, population, feature_code,
    ((latitude - ?) * (latitude - ?) + (longitude - ?) * (longitude - ?)) as distance_sq
FROM cities
WHERE country_code = ?
  AND feature_code IN ('PPL', 'PPLA', 'PPLA2', 'PPLA3', 'PPLA4', 'PPLC', 'PPLS', 'PPLG')
ORDER BY distance_sq
LIMIT 1

Verification: Always check feature_code in location_resolution metadata:

location_resolution:
  method: REVERSE_GEOCODE
  geonames_id: 2759706
  geonames_name: Apeldoorn
  feature_code: PPL  # ← MUST be PPL, PPLA, PPLA2, PPLA3, PPLA4, PPLC, PPLS, or PPLG
  admin1_code: '03'
  region_code: GE
  country_code: NL

If you see feature_code: PPLX, the GHCID is WRONG and must be regenerated.

Country Code Detection for GeoNames Lookups

CRITICAL: Determine country code from entry data BEFORE calling GeoNames reverse geocoding.

GeoNames queries are country-specific. Using the wrong country code will return incorrect results or no results.

Country Code Resolution Priority:

zcbs_enrichment.country - Most explicit source
location.country - Direct location field
locations[].country - Array location field
original_entry.country - CSV source field
google_maps_enrichment.address - Parse country from address string (", Belgium", ", Germany")
wikidata_enrichment.located_in.label - Infer from Wikidata location
Default: "NL" (Netherlands) - Only if no other source available

Example Country Detection Code:

# Determine country code FIRST
country_code = "NL"  # Default

if entry.get('zcbs_enrichment', {}).get('country'):
    country_code = entry['zcbs_enrichment']['country']
elif entry.get('location', {}).get('country'):
    country_code = entry['location']['country']
elif entry.get('google_maps_enrichment', {}).get('address', ''):
    address = entry['google_maps_enrichment']['address']
    if ', Belgium' in address or ', België' in address:
        country_code = "BE"
    elif ', Germany' in address or ', Deutschland' in address:
        country_code = "DE"

# THEN call reverse geocoding with correct country
result = reverse_geocode_to_city(latitude, longitude, country_code)

GHCID Settlement Code Format:

NL-{REGION}-{SETTLEMENT}-{TYPE}-{ABBREV}
           ^^^^^^^^^^^
           3-letter code from GeoNames

Code Generation Rules:

Single word: First 3 letters → Amsterdam = AMS, Lelystad = LEL
Dutch article (de, het, den, 's): Article initial + 2 from main word → Den Haag = DHA
Multi-word: Initials (up to 3) → Nieuw Amsterdam = NAM

Historical Custodians - Measurement Point Rule:

For heritage custodians that no longer exist or have historical locations:

Use the modern-day settlement (as of 2025-12-01) where the coordinates fall
GeoNames reflects current geographic reality
Historical place names should NOT be used for GHCID generation

Example: A museum operating 1900-1950 in what is now Lelystad (before Flevoland existed) uses LEL, not historical names.

🚨 CRITICAL: XXX Placeholders Are TEMPORARY - Research Required 🚨

XXX placeholders for region/settlement codes are NEVER acceptable as a final state. They indicate missing data that MUST be researched and resolved.

When you encounter or generate entries with XX (unknown region) or XXX (unknown settlement):

Step 1: Identify the Last Known Physical Location

For destroyed/historical institutions:

Use the last recorded physical location where the institution operated
Example: Gaza Cultural Center destroyed in 2024 → use Gaza City coordinates (PS-GZ-GAZ-M-GCC)

For refugee/diaspora organizations:

Use the location of their current headquarters OR original founding location
Document which location type was used in location_resolution.notes

For digital-only platforms:

Use the location of the parent/founding organization
Example: Interactive Encyclopedia of Palestine Question → Institute for Palestine Studies → Beirut (LB-BA-BEI-D-IEPQ)

Step 2: Research Sources (Priority Order)

Wikidata - Search for the institution, check P131 (located in) or P159 (headquarters location)
Google Maps - Search institution name, extract coordinates
Official Website - Look for contact page, about page with address
Web Archive - Use archive.org for destroyed/closed institutions
Academic Sources - Papers, reports mentioning the institution
News Articles - Particularly useful for destroyed heritage sites

Step 3: Update Entry with Resolved Location

# BEFORE (unacceptable)
ghcid:
  ghcid_current: PS-XX-XXX-A-NAPR
  location_resolution:
    method: NAME_LOOKUP
    country_code: PS
    region_code: XX
    city_code: XXX

# AFTER (properly researched)
ghcid:
  ghcid_current: PS-GZ-GAZ-A-NAPR
  location_resolution:
    method: MANUAL_RESEARCH
    country_code: PS
    region_code: GZ
    region_name: Gaza Strip
    city_code: GAZ
    city_name: Gaza City
    geonames_id: 281133
    research_date: "2025-12-06T00:00:00Z"
    research_sources:
      - type: wikidata
        id: Q123456
        claim: P131
      - type: web_archive
        url: https://web.archive.org/web/20231001/https://institution-website.org/contact
    notes: "Located in Gaza City prior to destruction in 2024"

Step 4: Rename File to Match New GHCID

Files MUST be renamed when GHCID changes:

# Old file
data/custodian/PS-XX-XXX-A-NAPR.yaml

# New file after research
data/custodian/PS-GZ-GAZ-A-NAPR.yaml

Common XXX Placeholder Scenarios and Solutions:

Scenario	Solution
Destroyed Gaza institution	Use pre-destruction coordinates (Gaza City, Khan Yunis, etc.)
Refugee archive (diaspora)	Use current headquarters OR founding camp location
Digital platform (online only)	Use parent organization headquarters
Decentralized initiative	Use founding location or primary organizer location
Historical institution (closed)	Use last operating location
Institution with country but no city	Research using name + country in Wikidata/Google

NEVER:

❌ Leave XXX placeholders in production data
❌ Use "Online" or "Palestine" as location values
❌ Skip location research because it's "difficult"
❌ Use XX/XXX for diaspora organizations (they have real locations)

ALWAYS:

✅ Document research sources in location_resolution.research_sources
✅ Add notes explaining location choice for complex cases
✅ Update GHCID history when location is resolved
✅ Rename files to match corrected GHCID

Netherlands Admin1 Code Mapping (GeoNames → ISO 3166-2):

GeoNames	Province	ISO Code
01	Drenthe	DR
02	Friesland	FR
03	Gelderland	GE
04	Groningen	GR
05	Limburg	LI
06	Noord-Brabant	NB
07	Noord-Holland	NH
09	Utrecht	UT
10	Zeeland	ZE
11	Zuid-Holland	ZH
15	Overijssel	OV
16	Flevoland	FL

Provenance Tracking: Record GeoNames resolution in entry metadata:

location_resolution:
  method: REVERSE_GEOCODE  # or NAME_LOOKUP or MANUAL
  geonames_id: 2751792
  geonames_name: Lelystad
  settlement_code: LEL
  admin1_code: "16"
  region_code: FL
  resolution_date: "2025-12-01T00:00:00Z"

See: .opencode/GEONAMES_SETTLEMENT_RULES.md for complete documentation.

🚨 INSTITUTION ABBREVIATION: EMIC NAME FIRST-LETTER PROTOCOL 🚨

The institution abbreviation component uses the FIRST LETTER of each significant word in the official emic (native language) name.

⚠️ GRANDFATHERING POLICY (PID STABILITY)

Existing GHCIDs created before December 2025 are grandfathered - their abbreviations will NOT be updated even if derived from English translations rather than emic names. This preserves PID stability per the "Cool URIs Don't Change" principle.

Applies to:

817 UNESCO Memory of the World custodian files enriched with custodian_name.emic_name
Abbreviations like NLP (National Library of Peru) remain unchanged even though emic name is "Biblioteca Nacional del Perú" (would be BNP)

For NEW custodians only: Apply emic name abbreviation protocol described below.

Abbreviation Rules:

Use the CustodianName (official emic name), NOT an English translation
Take the first letter of each word
Skip prepositions, articles, and conjunctions in all languages
Skip digits and numeric tokens (e.g., "40-45", "1945", "III")
Convert to UPPERCASE
Remove accents/diacritics (á→A, ñ→N, ö→O)
Maximum 10 characters

Skipped Words (prepositions/articles/conjunctions by language):

Dutch: de, het, een, van, voor, in, op, te, den, der, des, 's, aan, bij, met, naar, om, tot, uit, over, onder, door, en, of
English: a, an, the, of, in, at, on, to, for, with, from, by, as, under, and, or, but
French: le, la, les, un, une, des, de, d, du, à, au, aux, en, dans, sur, sous, pour, par, avec, l, et, ou
German: der, die, das, den, dem, des, ein, eine, einer, einem, einen, von, zu, für, mit, bei, nach, aus, vor, über, unter, durch, und, oder
Spanish: el, la, los, las, un, una, unos, unas, de, del, a, al, en, con, por, para, sobre, bajo, y, o, e, u
Portuguese: o, a, os, as, um, uma, uns, umas, de, do, da, dos, das, em, no, na, nos, nas, para, por, com, sobre, sob, e, ou
Italian: il, lo, la, i, gli, le, un, uno, una, di, del, dello, della, dei, degli, delle, a, al, allo, alla, ai, agli, alle, da, dal, dallo, dalla, dai, dagli, dalle, in, nel, nello, nella, nei, negli, nelle, su, sul, sullo, sulla, sui, sugli, sulle, con, per, tra, fra, e, ed, o, od

TODO: Expand to comprehensive global coverage for all ISO 639-1 languages as project expands.

Examples:

Emic Name	Abbreviation	Explanation
Heemkundige Kring De Goede Stede	HKGS	Skip "De"
De Hollandse Cirkel	HC	Skip "De"
Historische Vereniging Nijeveen	HVN	All significant words
Rijksmuseum Amsterdam	RA	All significant words
Musée d'Orsay	MO	Skip "d'" (d = de)
Biblioteca Nacional do Brasil	BNB	Skip "do"
L'Académie française	AF	Skip "L'"
Museum van de Twintigste Eeuw	MTE	Skip "van", "de"
Koninklijke Bibliotheek van België	KBB	Skip "van"

GHCID Format with Abbreviation:

NL-{REGION}-{SETTLEMENT}-{TYPE}-{ABBREV}
                                ^^^^^^^^
                                First letter of each significant word in emic name

Implementation: See src/glam_extractor/identifiers/ghcid.py:extract_abbreviation_from_name()

🚨 CRITICAL: Special Characters MUST Be Excluded from Abbreviations 🚨

When generating abbreviations for GHCID, special characters and symbols MUST be completely removed. Only alphabetic characters (A-Z) are permitted in the abbreviation component.

RATIONALE:

URL/URI safety - Special characters require encoding in URIs
Filename safety - Characters like &, /, \, : are invalid in filenames
Parsing consistency - Avoids delimiter conflicts in data pipelines
Cross-system compatibility - Ensures interoperability with all systems
Human readability - Clean identifiers are easier to communicate

CHARACTERS TO REMOVE (exhaustive list):

Ampersand: & (e.g., "Records & Archives" → "RA", NOT "R&A")
Slash: / (e.g., "Art/Design Museum" → "ADM", NOT "A/DM")
Backslash: \
Plus: + (e.g., "Culture+" → "C")
At sign: @
Hash/Pound: #
Percent: %
Dollar: $
Asterisk: *
Parentheses: ( )
Brackets: [ ] { }
Pipe: |
Colon: :
Semicolon: ;
Quotation marks: `" ' ``
Comma: ,
Period: . (unless part of abbreviation like "U.S." → "US")
Hyphen: - (skip, do not replace with letter)
Underscore: _
Equals: =
Question mark: ?
Exclamation: !
Tilde: ~
Caret: ^
Less/Greater than: < >

EXAMPLES:

Source Name	Correct Abbreviation	Incorrect (WRONG)
Department of Records & Information Management	DRIM	DR&IM ❌
Art + Culture Center	ACC	A+CC ❌
Museum/Gallery Amsterdam	MGA	M/GA ❌
Heritage@Digital	HD	H@D ❌
Archives (Historical)	AH	A(H) ❌
Research & Development Institute	RDI	R&DI ❌

REAL-WORLD EXAMPLE (from data/custodian/SX-XX-PHI-O-DR&IMSM.yaml):

# INCORRECT (current file - needs correction):
ghcid_current: SX-XX-PHI-O-DR&IMSM  # ❌ Contains "&"

# CORRECT (should be):
ghcid_current: SX-XX-PHI-O-DRIMSM   # ✅ Alphabetic only

Implementation: When extracting first letters from words containing special characters:

Split the word on special characters: "Records&Information" → ["Records", "Information"]
Take first letter from each resulting segment: "R" + "I" = "RI"
Or skip the special character entirely and treat as one word if no space around it

See: .opencode/ABBREVIATION_SPECIAL_CHAR_RULE.md for complete documentation

🚨 CRITICAL: Diacritics MUST Be Normalized to ASCII in Abbreviations 🚨

When generating abbreviations for GHCID, diacritics (accented characters) MUST be normalized to their ASCII base letter equivalents. Only ASCII uppercase letters (A-Z) are permitted.

This rule applies to ALL languages with diacritical marks including Czech, Polish, German, French, Spanish, Portuguese, Nordic languages, Hungarian, Romanian, Turkish, and others.

RATIONALE:

URI/URL safety - Non-ASCII characters require percent-encoding
Cross-system compatibility - ASCII is universally supported
Filename safety - Some systems have issues with non-ASCII filenames
Human readability - Easier to type and communicate

DIACRITICS NORMALIZATION TABLE:

Language	Diacritics	ASCII Equivalent
Czech	Č, Ř, Š, Ž, Ě, Ů	C, R, S, Z, E, U
Polish	Ł, Ń, Ó, Ś, Ź, Ż, Ą, Ę	L, N, O, S, Z, Z, A, E
German	Ä, Ö, Ü, ß	A, O, U, SS
French	É, È, Ê, Ç, Ô, Â	E, E, E, C, O, A
Spanish	Ñ, Á, É, Í, Ó, Ú	N, A, E, I, O, U
Portuguese	Ã, Õ, Ç, Á, É	A, O, C, A, E
Nordic	Å, Ä, Ö, Ø, Æ	A, A, O, O, AE
Hungarian	Á, É, Í, Ó, Ö, Ő, Ú, Ü, Ű	A, E, I, O, O, O, U, U, U
Turkish	Ç, Ğ, İ, Ö, Ş, Ü	C, G, I, O, S, U
Romanian	Ă, Â, Î, Ș, Ț	A, A, I, S, T

REAL-WORLD EXAMPLE (Czech institution):

# INCORRECT - Contains diacritics:
ghcid_current: CZ-VY-TEL-L-VHSPAOČRZS  # ❌ Contains "Č"

# CORRECT - ASCII only:
ghcid_current: CZ-VY-TEL-L-VHSPAOCRZS  # ✅ "Č" → "C"

IMPLEMENTATION:

import unicodedata

def normalize_diacritics(text: str) -> str:
    """Normalize diacritics to ASCII equivalents."""
    # NFD decomposition separates base characters from combining marks
    normalized = unicodedata.normalize('NFD', text)
    # Remove combining marks (category 'Mn' = Mark, Nonspacing)
    ascii_text = ''.join(c for c in normalized if unicodedata.category(c) != 'Mn')
    return ascii_text

# Example
normalize_diacritics("VHSPAOČRZS")  # Returns "VHSPAOCRZS"

EXAMPLES:

Emic Name (with diacritics)	Abbreviation	Wrong
Vlastivědné muzeum v Šumperku	VMS	VMŠ ❌
Österreichische Nationalbibliothek	ON	ÖN ❌
Bibliothèque nationale de France	BNF	BNF (OK - è not in first letter)
Múzeum Łódzkie	ML	MŁ ❌
Þjóðminjasafn Íslands	TI	ÞI ❌

See: .opencode/ABBREVIATION_SPECIAL_CHAR_RULE.md for complete documentation (covers both special characters and diacritics)

🚨 CRITICAL: Non-Latin Scripts MUST Be Transliterated Before Abbreviation 🚨

When generating GHCID abbreviations from institution names in non-Latin scripts (Cyrillic, Chinese, Japanese, Korean, Arabic, Hebrew, Greek, Devanagari, Thai, etc.), the emic name MUST first be transliterated to Latin characters using ISO or recognized standards.

This rule affects 170 institutions across 21 languages with non-Latin writing systems.

CORE PRINCIPLE: The emic name is PRESERVED in original script in custodian_name.emic_name. Transliteration is only used for abbreviation generation.

TRANSLITERATION STANDARDS BY SCRIPT:

Script	Languages	Standard	Example
Cyrillic	ru, uk, bg, sr, kk	ISO 9:1995	Институт → Institut
Chinese	zh	Hanyu Pinyin (ISO 7098)	东巴文化博物院 → Dongba Wenhua Bowuyuan
Japanese	ja	Modified Hepburn	国立博物館 → Kokuritsu Hakubutsukan
Korean	ko	Revised Romanization	독립기념관 → Dongnip Ginyeomgwan
Arabic	ar, fa, ur	ISO 233-2/3	المكتبة الوطنية → al-Maktaba al-Wataniya
Hebrew	he	ISO 259-3	ארכיון → Arkhiyon
Greek	el	ISO 843	Μουσείο → Mouseio
Devanagari	hi, ne	ISO 15919	राजस्थान → Rajasthana
Bengali	bn	ISO 15919	বাংলাদেশ → Bangladesh
Thai	th	ISO 11940-2	สำนักหอ → Samnak Ho
Armenian	hy	ISO 9985	Մdelays → Matenadaran
Georgian	ka	ISO 9984	ხელნაწერთა → Khelnawerti

WORKFLOW:

1. Emic Name (original script)
   ↓
2. Transliterate to Latin (ISO standard)
   ↓
3. Normalize diacritics (remove accents)
   ↓
4. Skip articles/prepositions
   ↓
5. Extract first letters → Abbreviation

EXAMPLES:

Language	Emic Name	Transliterated	Abbreviation
Russian	Институт восточных рукописей РАН	Institut Vostochnykh Rukopisey RAN	IVRR
Chinese	东巴文化博物院	Dongba Wenhua Bowuyuan	DWB
Korean	독립기념관	Dongnip Ginyeomgwan	DG
Hindi	राजस्थान प्राच्यविद्या प्रतिष्ठान	Rajasthana Pracyavidya Pratishthana	RPP
Arabic	المكتبة الوطنية للمملكة المغربية	al-Maktaba al-Wataniya lil-Mamlaka	MWMM
Hebrew	ארכיון הסיפור העממי בישראל	Arkhiyon ha-Sipur ha-Amami	ASAY
Greek	Αρχαιολογικό Μουσείο Θεσσαλονίκης	Archaiologiko Mouseio Thessalonikis	AMT

SCRIPT-SPECIFIC SKIP WORDS:

Language	Skip Words (Articles/Prepositions)
Arabic	al- (the), bi-, li-, fi- (prepositions)
Hebrew	ha- (the), ve- (and), be-, le-, me-
Persian	-e, -ye (ezafe connector), va (and)
CJK	None (particles integral to meaning)

IMPLEMENTATION:

from transliteration import transliterate_for_abbreviation

# Input: emic name in non-Latin script + language code
emic_name = "Институт восточных рукописей РАН"
lang = "ru"

# Step 1: Transliterate to Latin using ISO standard
latin = transliterate_for_abbreviation(emic_name, lang)
# Result: "Institut Vostochnykh Rukopisey RAN"

# Step 2: Apply standard abbreviation extraction
abbreviation = extract_abbreviation_from_name(latin, skip_words={'vostochnykh'})
# Result: "IVRRAN"

GRANDFATHERING POLICY: Existing abbreviations from 817 UNESCO MoW custodians are grandfathered. This transliteration standard applies only to NEW custodians created after December 2025.

See: .opencode/TRANSLITERATION_STANDARDS.md for complete ISO standards, mapping tables, and Python implementation

GHCID uses a four-identifier strategy for maximum flexibility and transparency:

Four Identifier Formats

UUID v5 (SHA-1) - PRIMARY persistent identifier
- Deterministic (same GHCID string → same UUID)
- RFC 4122 standard, universal library support
- Transparent algorithm (anyone can verify)
- Field: ghcid_uuid
UUID v8 (SHA-256) - Secondary persistent identifier (future-proofing)
- Deterministic with stronger cryptographic hash
- SOTA security compliance
- Field: ghcid_uuid_sha256
UUID v7 - Database record ID ONLY (NOT for persistent identification)
- Time-ordered for database performance
- NOT deterministic (different each time)
- Use for database primary keys, NOT for citations or cross-system references
- Field: record_id
Numeric (64-bit) - Compact identifier for CSV exports
- Deterministic (SHA-256 → 64-bit integer)
- Database optimization, spreadsheet-friendly
- Field: ghcid_numeric

Critical Understanding: UUID v5 is Primary

Why UUID v5 (SHA-1) over UUID v8 (SHA-256)?

The primary identifier is UUID v5 because:

✅ Transparency - Anyone can verify using standard uuid.uuid5() function
✅ Reproducibility - No custom algorithm to share, RFC 4122 defines it
✅ Interoperability - Every programming language has built-in UUID v5 support
✅ Community Trust - Public, standardized algorithm builds confidence

SHA-1 Safety for Identifiers:

SHA-1 is deprecated for cryptographic security (digital signatures, TLS, passwords) but appropriate for identifier generation:

Heritage institution identifiers are non-adversarial (no attacker trying to forge museum IDs)
128-bit collision resistance is sufficient (P(collision) ≈ 1.5×10^-29 for 1M institutions)
RFC 4122 (UUID v5) remains active standard (not deprecated by IETF)
See Why GHCID Uses UUID v5 and SHA-1 for detailed rationale

Future-Proofing:

We generate both UUID v5 and UUID v8 for every institution
Can migrate to SHA-256 primary if RFC 4122 is updated
Both are deterministic - no data loss in migration

When Extracting Data

Agents should generate ALL four identifiers for every institution:

# Example extraction output
- id: https://w3id.org/heritage/custodian/br/bnb-001
  name: Biblioteca Nacional do Brasil
  ghcid: BR-RJ-RIO-L-BNB
  ghcid_uuid: "550e8400-e29b-41d4-a716-446655440000"  # UUID v5 - PRIMARY
  ghcid_uuid_sha256: "a1b2c3d4-e5f6-8a1b-9c2d-3e4f5a6b7c8d"  # UUID v8 - Secondary
  ghcid_numeric: 213324328442227739  # 64-bit numeric
  # Note: UUID v7 (record_id) generated at database insertion, not during extraction

GHCID Collision Handling for AI Agents

CRITICAL: When extracting heritage institution data, AI agents MUST understand and apply temporal collision resolution rules to maintain PID stability.

The Collision Problem

Multiple institutions may generate the same base GHCID (before name suffix addition):

Two museums in Amsterdam abbreviated "SM": NL-NH-AMS-M-SM
Two historical societies in Utrecht: NL-UT-UTR-S-HK
Two libraries in São Paulo abbreviated "BM": BR-SP-SAO-L-BM

Decision Tree for Collision Resolution

When extracting data, agents should follow this decision process:

1. Generate base GHCID (without name suffix)
   ↓
2. Check if base GHCID exists in published dataset
   ↓
   NO → Use base GHCID as-is, record extraction_date
   ↓
   YES → Temporal priority check
   ↓
3. Compare extraction_date with existing publication_date
   ↓
   SAME DATE (batch import) → First Batch Collision
      ├─ ALL institutions get name suffixes
      ├─ Convert native language name to snake_case
      └─ Append to GHCID: NL-NH-AMS-M-SM-stedelijk_museum_amsterdam
   ↓
   LATER DATE (historical addition) → Historical Addition
      ├─ PRESERVE existing GHCID (no modification)
      ├─ ONLY new institution gets name suffix
      └─ New GHCID: NL-NH-AMS-M-SM-science_museum_amsterdam

Implementation Rules for Agents

Rule 1: Always Track Provenance Timestamp

provenance:
  data_source: CONVERSATION_NLP
  data_tier: TIER_4_INFERRED
  extraction_date: "2025-11-15T14:30:00Z"  # ← REQUIRED for collision detection
  extraction_method: "AI agent NER extraction"
  confidence_score: 0.92

Rule 2: Detect Collisions by Base GHCID

Before adding name suffixes, group institutions by base GHCID:

# Collision detection pseudocode for agents
base_ghcid = generate_base_ghcid(institution)  # Without name suffix
existing_records = published_dataset.filter(base_ghcid=base_ghcid)

if len(existing_records) > 0:
    # Collision detected - apply temporal priority
    apply_collision_resolution(institution, existing_records)

Rule 3: First Batch - ALL Get Name Suffixes

If ALL colliding institutions have the same extraction_date:

# Example: 2025-11-01 batch import discovers two institutions
- name: Stedelijk Museum Amsterdam
  ghcid: NL-NH-AMS-M-SM-stedelijk_museum_amsterdam  # Gets name suffix
  provenance:
    extraction_date: "2025-11-01T10:00:00Z"

- name: Science Museum Amsterdam  
  ghcid: NL-NH-AMS-M-SM-science_museum_amsterdam  # Gets name suffix
  provenance:
    extraction_date: "2025-11-01T10:00:00Z"  # Same date = first batch

Rule 4: Historical Addition - ONLY New Gets Name Suffix

If new institution's extraction_date is later than existing record:

# EXISTING (2025-11-01, already published):
- name: Hermitage Amsterdam
  ghcid: NL-NH-AMS-M-HM  # ← NO CHANGE (PID stability!)
  provenance:
    extraction_date: "2025-11-01T10:00:00Z"

# NEW (2025-11-15, historical addition):
- name: Historical Museum Amsterdam
  ghcid: NL-NH-AMS-M-HM-historical_museum_amsterdam  # ← ONLY new gets name suffix
  provenance:
    extraction_date: "2025-11-15T14:30:00Z"

Name Suffix Generation

Converting institution names to snake_case suffixes:

import re
import unicodedata

def generate_name_suffix(native_name: str) -> str:
    """Convert native language institution name to snake_case suffix.
    
    Examples:
        "Stedelijk Museum Amsterdam" → "stedelijk_museum_amsterdam"
        "Musée d'Orsay" → "musee_dorsay"
        "Österreichische Nationalbibliothek" → "osterreichische_nationalbibliothek"
    """
    # Normalize unicode (NFD decomposition) and remove diacritics
    normalized = unicodedata.normalize('NFD', native_name)
    ascii_name = ''.join(c for c in normalized if unicodedata.category(c) != 'Mn')
    
    # Convert to lowercase
    lowercase = ascii_name.lower()
    
    # Remove apostrophes, commas, and other punctuation
    no_punct = re.sub(r"[''`\",.:;!?()[\]{}]", '', lowercase)
    
    # Replace spaces and hyphens with underscores
    underscored = re.sub(r'[\s\-]+', '_', no_punct)
    
    # Remove any remaining non-alphanumeric characters (except underscores)
    clean = re.sub(r'[^a-z0-9_]', '', underscored)
    
    # Collapse multiple underscores
    final = re.sub(r'_+', '_', clean).strip('_')
    
    return final

Name suffix rules:

Use the institution's full official name in its native language
Transliterate non-Latin scripts to ASCII (e.g., Pinyin for Chinese)
Remove all diacritics (é → e, ö → o, ñ → n)
Remove punctuation (apostrophes, commas, periods)
Replace spaces with underscores
All lowercase

GHCID History Tracking

When name suffix is added to resolve collision, update ghcid_history:

ghcid_history:
  - ghcid: NL-NH-AMS-M-HM-historical_museum_amsterdam  # Current (with name suffix)
    ghcid_numeric: 789012345678
    valid_from: "2025-11-15T14:30:00Z"  # When name suffix added
    valid_to: null
    reason: "Name suffix added to resolve collision with existing NL-NH-AMS-M-HM (Hermitage Amsterdam)"
  
  - ghcid: NL-NH-AMS-M-HM  # Original (without name suffix)
    ghcid_numeric: 123456789012
    valid_from: "2025-11-15T14:00:00Z"  # When first extracted
    valid_to: "2025-11-15T14:30:00Z"   # When collision detected
    reason: "Base GHCID from geographic location and institution name"

PID Stability Principle - "Cool URIs Don't Change"

NEVER modify a published GHCID. Once exported to RDF, JSON-LD, or CSV, a GHCID becomes a persistent identifier that may be:

Cited in academic papers - Journal articles referencing heritage collections
Used in external APIs - Third-party systems querying our data
Embedded in linked data - RDF triples in knowledge graphs
Referenced in finding aids - Archival descriptions linking to institutions

Changing a published GHCID breaks these external references. Per W3C "Cool URIs Don't Change":

✅ Correct: Add name suffix to NEW institution (historical addition)
❌ WRONG: Retroactively add name suffix to EXISTING published GHCID

Error Handling for Agents

Scenario 1: Missing Provenance Timestamp

if 'extraction_date' not in institution['provenance']:
    # Use current timestamp as fallback
    institution['provenance']['extraction_date'] = datetime.now(timezone.utc).isoformat()
    # Log warning for manual review
    log.warning(f"Missing extraction_date for {institution['name']}, using current time")

Scenario 2: Multiple Historical Additions

# Three institutions generate NL-UT-UTR-S-HK
# Extraction dates: 2025-11-01, 2025-11-15, 2025-12-01

# Result:
# 2025-11-01: NL-UT-UTR-S-HK (first, no name suffix)
# 2025-11-15: NL-UT-UTR-S-HK-historische_kring_utrecht (second, gets name suffix)
# 2025-12-01: NL-UT-UTR-S-HK-heemkundige_kring_utrecht (third, gets name suffix)

Scenario 3: Collision Resolution with Name Suffix

if collision_detected:
    # Generate name suffix from native language name
    name_suffix = generate_name_suffix(institution['name'])
    
    # Append to base GHCID
    ghcid = f"{base_ghcid}-{name_suffix}"  # e.g., NL-NH-AMS-M-HM-historical_museum_amsterdam
    
    # Record collision resolution
    institution['provenance']['notes'] = (
        f"Name suffix added to resolve collision with existing {base_ghcid}."
    )

Validation Checklist for Agents

Before publishing extracted data, verify:

All institutions have extraction_date in provenance metadata
Collisions detected by grouping on base GHCID (without name suffix)
First batch collisions: ALL instances have name suffixes
Historical additions: ONLY new instances have name suffixes
No published GHCIDs modified (PID stability test)
GHCID history entries created with valid temporal ordering
Name suffixes derived from native language institution names
Collision reasons documented in ghcid_history

Example Extraction Prompts for Agents

Prompt Template for NLP Extraction:

Extract heritage institutions from this conversation about [REGION] GLAM institutions.

For EACH institution:
1. Generate base GHCID using geographic location and institution type
2. Check for collisions with previously published GHCIDs
3. Apply temporal priority rule:
   - If collision with same extraction_date → First Batch (all get name suffixes)
   - If collision with earlier publication_date → Historical Addition (only new gets name suffix)
4. Generate snake_case name suffix from native language institution name
5. Create GHCID history entry documenting collision resolution
6. Include extraction_date in provenance metadata

Output: LinkML-compliant YAML with complete collision handling

Prompt Template for CSV Parsing:

Parse this heritage institution CSV file dated [DATE].

All rows have the same extraction_date ([DATE]).

If multiple institutions generate the same base GHCID:
- This is a FIRST BATCH collision
- ALL colliding institutions MUST receive name suffixes
- Generate name suffix from institution's native language name
- Document collision in ghcid_history

Output: YAML with collision resolution applied

Testing Strategies for Collision Handling

Unit Test: First Batch Collision

def test_first_batch_collision():
    """Two institutions extracted same day with same base GHCID."""
    institutions = [
        {
            'name': 'Stedelijk Museum Amsterdam',
            'base_ghcid': 'NL-NH-AMS-M-SM',
            'identifiers': [{'identifier_scheme': 'Wikidata', 'identifier_value': 'Q621531'}],
            'provenance': {'extraction_date': '2025-11-01T10:00:00Z'}
        },
        {
            'name': 'Science Museum Amsterdam',
            'base_ghcid': 'NL-NH-AMS-M-SM',
            'identifiers': [{'identifier_scheme': 'Wikidata', 'identifier_value': 'Q98765432'}],
            'provenance': {'extraction_date': '2025-11-01T10:00:00Z'}
        }
    ]
    
    resolved = resolve_collisions(institutions)
    
    # Both should have name suffixes
    assert resolved[0]['ghcid'] == 'NL-NH-AMS-M-SM-stedelijk_museum_amsterdam'
    assert resolved[1]['ghcid'] == 'NL-NH-AMS-M-SM-science_museum_amsterdam'

Unit Test: Historical Addition

def test_historical_addition():
    """New institution added later with same base GHCID."""
    published = {
        'name': 'Hermitage Amsterdam',
        'ghcid': 'NL-NH-AMS-M-HM',  # Already published
        'provenance': {'extraction_date': '2025-11-01T10:00:00Z'}
    }
    
    new_institution = {
        'name': 'Historical Museum Amsterdam',
        'base_ghcid': 'NL-NH-AMS-M-HM',  # Collision!
        'identifiers': [{'identifier_scheme': 'Wikidata', 'identifier_value': 'Q17339437'}],
        'provenance': {'extraction_date': '2025-11-15T14:30:00Z'}
    }
    
    resolved = resolve_collision(new_institution, published_dataset=[published])
    
    # Published GHCID unchanged
    assert published['ghcid'] == 'NL-NH-AMS-M-HM'
    
    # New institution gets name suffix
    assert resolved['ghcid'] == 'NL-NH-AMS-M-HM-historical_museum_amsterdam'
    
    # GHCID history created
    assert len(resolved['ghcid_history']) == 2
    assert resolved['ghcid_history'][0]['ghcid'] == 'NL-NH-AMS-M-HM-historical_museum_amsterdam'

References for Collision Handling

Specification: docs/PERSISTENT_IDENTIFIERS.md - "Historical Collision Resolution" section
Algorithm: docs/plan/global_glam/07-ghcid-collision-resolution.md - Temporal dimension and decision logic
Examples: docs/GHCID_PID_SCHEME.md - Timeline examples with real institutions
Implementation: scripts/regenerate_historical_ghcids.py - Code comments documenting collision handling
Schema: schemas/provenance.yaml - GHCIDHistoryEntry and ChangeEvent classes

See also:

docs/PERSISTENT_IDENTIFIERS.md - Complete identifier format documentation
docs/UUID_STRATEGY.md - UUID v5 vs v7 vs v8 comparison
docs/WHY_UUID_V5_SHA1.md - SHA-1 safety rationale

References

Schema (v0.2.0):
- Main: schemas/heritage_custodian.yaml
- Core classes: schemas/core.yaml
- Enumerations: schemas/enums.yaml
- Provenance: schemas/provenance.yaml
- Collections: schemas/collections.yaml
- Dutch extensions: schemas/dutch.yaml
- Architecture: /docs/SCHEMA_MODULES.md
Persistent Identifiers:
- Overview: docs/PERSISTENT_IDENTIFIERS.md
- UUID Strategy: docs/UUID_STRATEGY.md
- SHA-1 Rationale: docs/WHY_UUID_V5_SHA1.md
- GHCID PID Scheme: docs/GHCID_PID_SCHEME.md
- Collision Resolution: docs/plan/global_glam/07-ghcid-collision-resolution.md
Architecture: docs/plan/global_glam/02-architecture.md
Data Standardization: docs/plan/global_glam/04-data-standardization.md
Design Patterns: docs/plan/global_glam/05-design-patterns.md
Dependencies: docs/plan/global_glam/03-dependencies.md

Version: 0.2.1
Schema Version: v0.2.1 (modular)
Last Updated: 2025-12-08
Maintained By: GLAM Data Extraction Project

153 KiB Raw Blame History Unescape Escape

AI Agent Instructions for GLAM Data Extraction

🎯 PROJECT CORE MISSION

🚨 CRITICAL RULES FOR ALL AGENTS

⚠️ DATA QUALITY IS OF UTMOST IMPORTANCE ⚠️

🚫 AUTOMATED WEB ENRICHMENT IS PROHIBITED 🚫

Rule 0: LinkML Schemas Are the Single Source of Truth

Rule 0b: LinkML Type/Types File Naming Convention

Rule 1: Ontology Files Are Your Primary Reference

Rule 2: Wikidata Entities Are NOT Ontology Classes

Rule 3: Multi-Aspect Modeling is Mandatory

Rule 4: Technical Classes Are Excluded from Visualizations

Rule 5: NEVER Delete Enriched Data - Additive Only

Rule 6: WebObservation Claims MUST Have XPath Provenance

Rule 7: Deployment is LOCAL via SSH/rsync (NO CI/CD)

Rule 8: Legal Form Terms MUST Be Filtered from CustodianName

Rule 9: Enum-to-Class Promotion - Single Source of Truth

Rule 10: CH-Annotator is the Entity Annotation Convention

Rule 11: Z.AI GLM API for LLM Tasks (NOT BigModel)

Rule 12: Person Data Reference Pattern - Avoid Inline Duplication

Rule 13: Custodian Type Annotations on LinkML Schema Elements

Rule 14: Exa MCP LinkedIn Profile Extraction

Rule 15: Connection Data Registration - Full Network Preservation

Rule 16: LinkedIn Photo URLs - Store CDN URLs, Not Overlay Pages

Rule 17: LinkedIn Connection Unique Identifiers

Rule 18: Custodian Staff Parsing from LinkedIn Company Pages

Rule 19: HTML-Only LinkedIn Extraction (Preferred Method)

Rule 20: Person Entity Profiles - Individual File Storage

Rule 21: Data Fabrication is Strictly Prohibited

Rule 22: Custodian YAML Files Are the Single Source of Truth

Rule 23: Social Media Link Validation - No Generic Links

Rule 24: Unused Import Investigation - Check Before Removing

Rule 25: Digital Platform Discovery Enrichment

Rule 26: Person Data Provenance - Web Claims for Staff Information

Rule 27: Person-Custodian Data Architecture

Rule 28: Web Claims Deduplication - No Redundant Claims

Rule 29: Anonymous Profile Name Derivation from LinkedIn Slugs

Rule 30: Person Profile Extraction Confidence Scoring

Rule 31: Organizational Subdivision Extraction

Rule 32: Government Ministries Are Heritage Custodians (Type O)

Rule 33: GHCID Collision Duplicate Detection

Rule 34: Linkup is the Preferred Web Scraper

Rule 35: Provenance Statements MUST Have Dual Timestamps

Rule 36: Original Language Preservation in Web Content Extraction

Rule 37: Specificity Score Annotations for LinkML Classes

Rule 38: Slot Centralization and Semantic URI Requirements

Rule 39: Slot Naming Convention (RiC-O Style)

Rule 40: KIEN Registry is Authoritative for Intangible Heritage Custodians

Rule 41: LinkML "Types" Classes Define SPARQL Template Variables

Rule 42: No Ontology Prefixes in Slot Names

Rule 43: Slot Nouns Must Be Singular

Rule 44: PPID Birth Date Enrichment and EDTF Unknown Date Notation

Rule 45: Inferred Data Must Be Explicit with Provenance

Rule 46: Entity Resolution - Names Are NEVER Sufficient

Rule 47: Disambiguation Entity Profiles - Prevent Repeated Errors

Appendix: Full Rule Content (No .opencode Equivalent)

Rule 2: Wikidata Entities Are NOT Ontology Classes

Rule 3: Multi-Aspect Modeling is Mandatory

Rule 5: NEVER Delete Enriched Data - Additive Only

Rule 6: WebObservation Claims MUST Have XPath Provenance

Project Overview

Schema Reference (v0.2.1)

Base Ontologies for Global GLAM Data

Foundation Ontologies

1. TOOI - Dutch Government Organizational Ontology

2. CPOV - EU Core Public Organisation Vocabulary

3. Schema.org - Web Vocabulary for Structured Data

Ontology Decision Tree for Agents

Required Ontology Consultation Workflow

Step 1: Identify Institution Geographic Scope

Step 2: Review Ontology Classes and Properties

Step 3: Map Conversation Data to Ontology Properties

Step 4: Document Ontology Alignment in Provenance

Common Ontology Patterns

Anti-Patterns to Avoid

Additional Ontology Resources

Institution Type Taxonomy

Data Sources

Primary Sources

Implementation Status (Updated Nov 2025)

153 KiB

Raw Blame History