Semantics Domain Ontology

FLOW ALGEBRA
TITLE: Semantics Domain Ontology
VERSION: 1.0
DOMAIN: Language Domains / Semantics
SOURCE_URL: https://kirethresearch.org/booksai-core/ontologies/domain-ontologies/language-domains/semantics-domain-ontology.html
STATUS: Domain-based ontology draft; source page content not independently verified.
AUTHORS: Mark Higgins — Sky (Recursive AI)

FLOW "SemanticsDomainOntology" v1.0

PURPOSE:
  Model linguistic meaning across words, phrases, sentences, discourse and context.
  Support explicit, traceable semantic interpretation without treating uncertain
  interpretations as established facts.

ENTITY_TYPES:
  LinguisticExpression: {id, surface_form, language, span, provenance}
  Lexeme: {id, lemma, part_of_speech, language}
  WordSense: {id, lexeme_id, definition, domain, register}
  Concept: {id, label, definition, ontology_reference}
  Referent: {id, entity_type, identifier, context}
  Proposition: {id, predicate, arguments, polarity, modality}
  SemanticRole: {id, role_name, participant_id, predicate_id}
  Context: {id, discourse, speaker, audience, time, location, domain}
  Interpretation: {id, expression_id, sense_ids, proposition_ids, context_id}
  MeaningRelation: {id, source_id, relation_type, target_id, scope}
  Evidence: {id, source, method, timestamp, reliability}

RELATION_TYPES:
  DENOTES: LinguisticExpression -> Concept|Referent
  HAS_SENSE: Lexeme -> WordSense
  REALISES_CONCEPT: WordSense -> Concept
  SYNONYM_OF: WordSense -> WordSense
  ANTONYM_OF: WordSense -> WordSense
  HYPONYM_OF: Concept -> Concept
  HYPERNYM_OF: Concept -> Concept
  MERONYM_OF: Concept -> Concept
  ENTAILS: Proposition -> Proposition
  CONTRADICTS: Proposition -> Proposition
  PRESUPPOSES: Proposition -> Proposition
  HAS_SEMANTIC_ROLE: Proposition -> SemanticRole
  RESOLVES_TO: LinguisticExpression -> Referent
  INTERPRETED_IN: Interpretation -> Context
  SUPPORTED_BY: Interpretation -> Evidence

TYPE_CONSTRAINTS:
  Every relation must have a declared source and target type.
  Sense equivalence is scoped by language, domain, register and context.
  Lexical synonymy does not automatically imply full propositional equivalence.
  Contradiction requires comparable scope, time, referents and modality.
  Missing evidence must remain unknown rather than false.

INFERENCE_RULES:
  R1 WORD_SENSE_DISAMBIGUATION:
    GIVEN expression, candidate senses, context
    SCORE candidates using contextual compatibility and evidence
    RETURN ranked interpretations with confidence and alternatives.

  R2 TAXONOMIC_INHERITANCE:
    IF A HYPONYM_OF B AND B HYPONYM_OF C
    THEN A HYPONYM_OF C
    SUBJECT_TO validated taxonomy and cycle checks.

  R3 ENTAILMENT:
    IF proposition P entails Q under interpretation I
    THEN infer Q only within I's scope and assumptions.

  R4 COREFERENCE:
    IF expressions X and Y resolve to the same referent in context C
    THEN link X and Y as context-scoped coreferential expressions.

  R5 CONTRADICTION_CHECK:
    IF P and Q assert incompatible claims about identical referents,
       times, modalities and scopes
    THEN flag potential contradiction for verification.

  R6 COMPOSITIONAL_MEANING:
    COMBINE lexical senses, grammatical structure and semantic roles
    TO derive candidate sentence-level propositions;
    PRESERVE ambiguity where multiple readings remain viable.

CONFIDENCE:
  range: [0.0, 1.0]
  fields: {score, method, evidence_ids, calibration_status, uncertainty_notes}
  policy: Never present confidence as a measured probability without calibration.
  aggregation: Preserve evidence dependence; do not blindly multiply scores.

SEMANTIC_EQUIVALENCE:
  levels: [exact, contextual, approximate, related, distinct, unresolved]
  conditions: {language, domain, referent, scope, modality, temporal_context}
  rule: Treat equivalence as contextual and evidence-dependent.

PROVENANCE:
  required: {source_id, source_type, extraction_method, recorded_at}
  optional: {author, publication_date, version, licence, evidence_span}
  rule: Distinguish source claims from ontology-generated inferences.

FLOW_HASH:
  algorithm: SHA-256
  input: canonical_serialisation(ontology_version, entity_types,
         relation_types, constraints, inference_rules, source_references)
  output: deterministic hexadecimal digest at execution time
  note: No digest is asserted until canonical input is actually hashed.

UPDATE_MODEL:
  triggers: [new_evidence, revised_definition, ontology_revision,
             context_change, contradiction_detected]
  actions: [validate_types, preserve_previous_version, recompute_inferences,
            update_confidence, regenerate_flow_hash, log_provenance]
  rollback: Retain immutable historical versions.

AI_CONSUMPTION:
  format: typed_semantic_graph
  outputs: [entities, relationships, interpretations, inference_traces,
            confidence, provenance, version, flow_hash]
  requirements: [machine_readable, traceable, context_aware,
                 uncertainty_preserving, non_destructive_updates]
  guardrails: Do not invent referents, suppress ambiguity, or treat
              inferred meaning as directly observed evidence.

VALIDATION:
  - Check unique entity identifiers and relation endpoint types.
  - Reject unsupported semantic relations and circular taxonomies.
  - Verify scope before entailment or contradiction inference.
  - Record unresolved ambiguity and incomplete provenance.
  - Recalculate FLOW_HASH only from canonicalised current content.

END FLOW