glam/data/entity_annotation/modules/hypernyms/agt.yaml
2025-12-05 15:30:23 +01:00

312 lines
12 KiB
YAML

# =============================================================================
# GLAM-NER: AGENT HYPERNYM MODULE
# =============================================================================
# Module: hypernyms/agt.yaml
# Parent: entity_annotation_rules_v1.7.0_unified.yaml
# Purpose: AGENT entity type - entities capable of intentional action
# =============================================================================
# BREAKING CHANGE v1.7.0: Renamed from BEING (BEI) to AGENT (AGT)
# Rationale: "Being" implies human-centric ontology. CIDOC-CRM E39_Actor is
# the proper hypernym for ALL entities capable of intentional action.
# =============================================================================
id: https://w3id.org/glam/ner/hypernym/agent
name: glam-ner-agent-hypernym
AGENT:
code: "AGT"
definition: |
Entities capable of intentional action. Includes humans (historical and
contemporary), animals, AI systems, fictional characters, mythological
figures, and collectives. The defining characteristic is AGENCY - the
capacity to act, make decisions, and bear responsibility.
This is the broadest actor class, encompassing all entities that can:
- Perform intentional actions (create, destroy, transfer, modify)
- Hold beliefs, desires, or goals
- Bear moral or legal responsibility
- Be attributed authorship or causation
design_rationale: |
CIDOC-CRM E39_Actor ("a persistent item that has the potential to perform
intentional actions") is the correct hypernym. The former "BEING" label
was anthropocentric, excluding valid actors like:
- AI systems creating art or making curatorial decisions
- Named animals with documented agency (working animals, famous pets)
- Fictional characters who are subjects of scholarly study
- Robots performing heritage conservation tasks
TEI P5 provides the Character (roleName) model for fictional entities.
FOAF provides Agent as a superclass of Person and Group.
# ---------------------------------------------------------------------------
# ONTOLOGY MAPPINGS
# ---------------------------------------------------------------------------
ontology_mappings:
primary_class: "crm:E39_Actor"
primary_class_definition: |
CIDOC-CRM E39 Actor: "This class comprises people, either individually as
members of groups or members of groups. A gathering of members of
E21 Persons becomes an instance of E74 Group when it exhibits collective
agency, that is, it can perform actions as a unit."
alternative_classes:
- "foaf:Agent"
- "schema:Thing" # schema:Person and schema:Organization are subclasses
- "prov:Agent"
linkml_mapping:
class_uri: "crm:E39_Actor"
exact_mappings:
- "foaf:Agent"
- "prov:Agent"
close_mappings:
- "schema:Person" # More specific, human-only
related_mappings:
- "dcterms:Agent"
nerd_class: "nerd:Person"
nerd_deprecation_note: |
DEPRECATED: NERD's Person class is too narrow. NERD was designed for
news/journalism NER where non-human agents are rare. For Digital
Humanities, CIDOC-CRM E39_Actor is authoritative.
Retain NERD mapping ONLY for NLP pipeline interchange, NOT as semantic authority.
pico_class: "picom:PersonObservation"
pico_note: |
In PiCO, textual mentions create PersonObservation instances linked
to reconstructed Person entities via picom:isObservationOf. This
observation/reconstruction pattern applies to ALL agent subcategories.
# ---------------------------------------------------------------------------
# SUBCATEGORIES
# ---------------------------------------------------------------------------
subcategories:
# ----- HUMAN AGENTS -----
PERSON:
code: "AGT.PER"
definition: "Individual human beings, historical or contemporary"
examples:
- "Rembrandt van Rijn"
- "Queen Beatrix"
- "Jan de Bakker"
- "Marie Curie"
ontology_class: "crm:E21_Person"
linkml_mapping:
class_uri: "crm:E21_Person"
exact_mappings:
- "foaf:Person"
- "schema:Person"
- "rico:Person"
STAFF:
code: "AGT.STF"
definition: "Personnel of heritage institutions in professional roles"
examples:
- "Dr. Maria van den Berg, Director"
- "Jan Pietersen, Curator of Prints"
- "Chief Archivist Emma de Vries"
ontology_class: "picom:PersonObservation"
org_ontology_mapping:
membership: "org:Membership"
role: "org:Role"
post: "org:Post"
note: |
Links to institution via org:memberOf or org:holds (for Posts).
The org:Membership class represents the n-ary relationship between
an Agent, an Organization, and a Role. Use org:Post when the position
exists independently of the person filling it.
For the ROLE itself (e.g., "Director", "Curator"), see ROLE hypernym.
# ----- COLLECTIVE AGENTS -----
COLLECTIVE:
code: "AGT.COL"
definition: |
Named collectives of agents acting as a unit but WITHOUT formal
organizational structure. For formal organizations, use GROUP hypernym.
examples:
- "The Dutch Masters"
- "The Impressionists"
- "The Founding Fathers"
- "Anonymous (hacker collective)"
ontology_class: "crm:E74_Group"
alternative_classes:
- "foaf:Group"
note: |
Collectives exhibit collective agency but lack:
- Legal personality
- Formal membership rules
- Organizational hierarchy
For formal organizations (museums, companies), use GROUP hypernym.
For informal project collaborations, use org:OrganizationalCollaboration.
# ----- FICTIONAL/MYTHOLOGICAL AGENTS -----
FICTIONAL:
code: "AGT.FIC"
definition: |
Characters from fiction, mythology, legend, or religious traditions
who are subjects of scholarly study or cultural analysis.
examples:
- "Sherlock Holmes"
- "Harry Potter"
- "Hamlet"
- "Don Quixote"
ontology_class: "crm:E21_Person"
alternative_classes:
- "tei:character" # TEI P5 character element
linkml_mapping:
class_uri: "crm:E21_Person"
exact_mappings: []
close_mappings:
- "schema:Person"
note: "Use crm:P2_has_type with value 'fictional' to distinguish"
tei_note: |
TEI P5 uses <person> with @role="fictional" or nests within
<listPerson type="fictional">. The <character> element (from
TEI Drama module) is more specific for dramatic personae.
MYTHOLOGICAL:
code: "AGT.MYT"
definition: |
Gods, deities, legendary figures, and supernatural beings from
religious or mythological traditions.
examples:
- "Apollo"
- "Thor"
- "Vishnu"
- "Anansi"
- "King Arthur"
- "Siegfried"
ontology_class: "crm:E21_Person"
note: |
Use crm:P2_has_type to indicate mythological/divine status.
Mythological figures may have historical cult/worship data
(temples, festivals) even though the entity is non-physical.
# ----- NON-HUMAN AGENTS -----
ANIMAL:
code: "AGT.ANI"
definition: |
Named individual animals with documented agency or cultural
significance. NOT species names (use THING.TAX for taxonomy).
examples:
- "Dolly the sheep (first cloned mammal)"
- "Jumbo the elephant"
- "Hachiko (famous loyal dog)"
- "Wojtek the soldier bear"
- "Paul the Octopus (World Cup predictor)"
ontology_class: "crm:E39_Actor"
note: |
Animals qualify as agents when they:
- Have individual names (not just species)
- Performed documented actions
- Have cultural/historical significance
Generic animal mentions ("a cat", "the horses") are NOT agents.
ARTIFICIAL:
code: "AGT.ART"
definition: |
Artificial agents: AI systems, robots, software agents, and
automated systems capable of autonomous decision-making or action.
examples:
- "DALL-E (AI image generator)"
- "AlphaGo (game-playing AI)"
- "Sophia (humanoid robot)"
- "Watson (IBM's AI system)"
- "GPT-4 (language model)"
ontology_class: "crm:E39_Actor"
alternative_classes:
- "prov:SoftwareAgent"
linkml_mapping:
class_uri: "crm:E39_Actor"
exact_mappings:
- "prov:SoftwareAgent"
note: |
PROV-O defines prov:SoftwareAgent as "A software agent is running
software." Use when the AI/robot is the proximate cause of an action,
distinct from the human programmers or operators.
note: |
Artificial agents are increasingly relevant for heritage:
- AI systems making curatorial decisions
- Robots performing conservation tasks
- Automated digitization systems
- AI-generated art and authorship questions
Attribution of agency to AI is context-dependent and evolving.
# ---------------------------------------------------------------------------
# INCLUSION RULES
# ---------------------------------------------------------------------------
inclusion_rules:
- id: "AGT_INC001"
rule: "Tag agent names even when only partial name appears"
examples:
- "Rembrandt (given name only)"
- "Van Gogh (surname only)"
- "GPT (abbreviated AI name)"
- id: "AGT_INC002"
rule: "Tag agents identified by title + name"
examples:
- "Professor Einstein"
- "Dr. Curie"
- "King Willem-Alexander"
- id: "AGT_INC003"
rule: "Tag staff members with their institutional context"
examples:
- "Director Jan de Wit"
- "Curator of Medieval Art at the Rijksmuseum"
- id: "AGT_INC004"
rule: "Tag named collectives acting as unified agents"
examples:
- "The Impressionists"
- "Anonymous"
- "The Beatles"
- id: "AGT_INC005"
rule: "Tag fictional/mythological characters when subjects of analysis"
examples:
- "Hamlet's soliloquy"
- "representations of Apollo"
- "Harry Potter merchandise"
- id: "AGT_INC006"
rule: "Tag AI/robot agents when attributed with actions"
examples:
- "art created by DALL-E"
- "AlphaGo defeated Lee Sedol"
# ---------------------------------------------------------------------------
# EXCLUSION RULES
# ---------------------------------------------------------------------------
exclusion_rules:
- id: "AGT_EXC001"
rule: "Do NOT tag generic role descriptions without names"
examples:
- "the curator (generic)"
- "a visitor (generic)"
- "staff members (generic plural)"
note: "For role terms themselves, see ROLE hypernym"
- id: "AGT_EXC002"
rule: "Do NOT tag pronouns"
examples:
- "he"
- "she"
- "they"
- "it"
- id: "AGT_EXC003"
rule: "Do NOT tag species names (use THING.TAX instead)"
examples:
- "elephants (species, not individual)"
- "Homo sapiens"
- "cats (generic)"
- id: "AGT_EXC004"
rule: "Do NOT tag tools or software without agency attribution"
examples:
- "Photoshop (tool, not agent)"
- "the database (system, not agent)"
note: "Only tag AI when it is the attributed actor, not just a tool used"