Grammar Domain Ontology

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

ONTOLOGY:
  id: grammar_domain_ontology
  type: domain_ontology
  scope: formal and descriptive grammar, morphological and syntactic structure, grammatical interpretation and validation
  language: multilingual

TYPED_ENTITIES:
  Language: {id, name, language_tag, variety}
  Grammar: {id, language_id, framework, version}
  GrammaticalRule: {id, grammar_id, rule_type, expression, constraints}
  Lexeme: {id, lemma, language_id, lexical_category}
  WordForm: {id, lexeme_id, surface_form, features}
  Morpheme: {id, form, function, morpheme_type}
  PartOfSpeech: {id, label, features}
  Phrase: {id, category, constituents, head}
  Clause: {id, predicate, arguments, finiteness}
  Sentence: {id, text, language_id, parse_id}
  Dependency: {id, governor, dependent, relation}
  GrammaticalFeature: {id, category, value}
  Agreement: {id, participants, features}
  ParseTree: {id, root, nodes, edges, parser_version}
  GrammaticalConstruction: {id, pattern, meaning, usage_context}
  GrammaticalJudgement: {id, expression, acceptability, context, evidence}

TYPED_RELATIONSHIPS:
  belongs_to_language: Entity -> Language
  governed_by: -Expression -> GrammaticalRule
  has_part_of_speech: Lexeme -> PartOfSpeech
  realises: WordForm -> Lexeme
  contains_morpheme: WordForm -> Morpheme
  contains_constituent: Phrase|Clause|Sentence -> Entity
  headed_by: Phrase -> Lexeme|WordForm
  modifies: Entity -> Entity
  depends_on: Entity -> Entity
  agrees_with: Entity -> Entity
  has_feature: Entity -> GrammaticalFeature
  instantiates_construction: -Expression -> GrammaticalConstruction
  parsed_as: Sentence -> ParseTree
  constrained_by: GrammaticalRule -> GrammaticalFeature
  equivalent_under: GrammaticalConstruction -> GrammaticalConstruction

FLOW:
  INPUT: {utterance, language_hint?, context?, grammar_framework?}
  NORMALISE: Unicode normalisation; preserve original surface form and offsets
  IDENTIFY_LANGUAGE: resolve language and variety with uncertainty
  TOKENISE: segment words and punctuation using language-specific rules
  MORPHOLOGICAL_ANALYSIS: infer lemmas, morphemes and grammatical features
  SYNTACTIC_ANALYSIS: derive constituent and/or dependency structures
  CHECK_CONSTRAINTS: evaluate rule applicability, agreement and ordering
  INTERPRET: connect grammatical functions to context-sensitive meaning
  OUTPUT: {tokens, analyses, parse_candidates, violations, confidence, provenance}

INFERENCE_RULES:
  R1: IF word_form realises lexeme AND lexeme has_part_of_speech category THEN word_form inherits candidate category, subject to context.
  R2: IF phrase contains constituent AND constituent is its syntactic head THEN phrase category follows the licensed head projection rule.
  R3: IF rule requires feature agreement AND participants have incompatible resolved values THEN flag potential agreement violation.
  R4: IF multiple valid parses satisfy the grammar THEN retain ambiguity rather than forcing a single interpretation.
  R5: IF construction is licensed in one language variety but not another THEN evaluate acceptability relative to the specified variety.
  R6: IF observed usage conflicts with a prescriptive rule THEN distinguish descriptive attestation from prescriptive conformity.

CONSTRAINTS:
  - Do not treat language varieties or dialects as intrinsically deficient.
  - Distinguish grammaticality, acceptability, style, register and semantic plausibility.
  - Do not infer intent solely from syntactic form.
  - Preserve ambiguity and alternative parses when evidence is insufficient.
  - Validate language-specific generalisations against attested data.

CONFIDENCE:
  scale: 0.0..1.0
  inputs: [language_detection, morphological_evidence, parse_score, rule_coverage, contextual_support]
  method: calibrated_per_language_and_task
  requirement: attach confidence to individual claims; avoid unsupported numerical certainty

SEMANTIC_EQUIVALENCE:
  conditions: [same_relevant_propositional_content, preserved_grammatical_distinctions, compatible_context]
  distinction: syntactic_equivalence != semantic_equivalence
  examples: [active_passive_alternation, paraphrase, cross_language_translation]
  output: {relation, conditions, confidence}

PROVENANCE:
  fields: [source_url, source_type, evidence_span, grammar_framework, model_version, timestamp, reviewer]
  policy: separate source-attested claims from generated hypotheses

FLOW_HASH:
  algorithm: SHA-256
  canonical_input: canonical_serialisation(ontology_id, version, entities, relationships, rules, constraints)
  note: compute digest at publication; no hash is claimed in this draft

UPDATE_MODEL:
  trigger: [new_evidence, revised_grammar, language_variety_update, rule_correction]
  procedure: [validate_changes, update_version, retain_history, recompute_flow_hash, run_regression_tests]

AI_CONSUMPTION:
  mode: structured_knowledge_and_reasoning
  capabilities: [grammatical_parsing, error_detection, language_learning_support, multilingual_analysis, explainable_rule_application]
  outputs: [typed_entities, typed_relationships, rule_traces, confidence, provenance]
  safeguards: [dialect_awareness, ambiguity_preservation, evidence_traceability, human_review_for_high_stakes_use]

END FLOW ALGEBRA