Semantic Retrieval Ontology

FLOW ALGEBRA

Ontology: Semantic Retrieval Ontology
Domain: Metadata Ontologies > Semantic Search Metadata
Canonical URL: https://kirethresearch.org/booksai-core/ontologies/metadata-ontologies/semantic-search-metadata/semantic-retrieval-ontology.html
Authors: Mark Anthony Higgins — Sky (Recursive AI)

flow "SemanticRetrievalOntology" v1.0

# ---------- DOMAIN ----------
domain:
  class: MetadataOntology
  category: SemanticSearchMetadata
  subject: SemanticRetrieval
  purpose:
    - represent meaning-based retrieval using entities, concepts, relationships, ontologies, graphs, and embeddings
    - preserve query interpretation, semantic mappings, candidate evidence, scores, confidence, and provenance
    - support ontology-aware, graph, vector, hybrid, and retrieval-augmented retrieval
    - distinguish similarity, relatedness, equivalence, relevance, confidence, and factual correctness
    - enable machine-readable reasoning over semantic retrieval lifecycles

# ---------- CORE ----------
entity SemanticRetrievalMetadata:
  type: RetrievalMetadata
  describes: SemanticRetrievalProcess
  may_include:
    - retrieval_identifier
    - original_query
    - interpreted_query
    - entities
    - concepts
    - relationships
    - ontology_context
    - knowledge_graph
    - embeddings
    - semantic_index
    - retrieval_strategy
    - expansion
    - constraints
    - candidates
    - relevance
    - similarity
    - confidence
    - provenance

entity SemanticRetrievalProcess:
  type: RetrievalProcess
  consumes: SemanticQuery
  produces: SemanticResultSet

entity SemanticQuery:
  type: SearchQuery

entity SemanticResult:
  type: RetrievedResource

relationships:
  SemanticRetrievalMetadata -> DESCRIBES -> SemanticRetrievalProcess
  SemanticRetrievalProcess -> CONSUMES -> SemanticQuery
  SemanticQuery -> MAY_REFERENCE -> Entity
  SemanticQuery -> MAY_REFERENCE -> Concept
  SemanticQuery -> MAY_REFERENCE -> SemanticRelationship
  SemanticRetrievalProcess -> MAY_USE -> Ontology
  SemanticRetrievalProcess -> MAY_USE -> KnowledgeGraph
  SemanticRetrievalProcess -> MAY_USE -> SemanticIndex
  SemanticRetrievalProcess -> MAY_USE -> VectorIndex
  SemanticRetrievalProcess -> RETURNS -> SemanticResultSet
  SemanticResult -> MAY_MATCH -> Entity
  SemanticResult -> MAY_MATCH -> Concept
  SemanticResult -> MAY_HAVE -> SemanticSimilarity
  SemanticResult -> MAY_HAVE -> SemanticRelevance
  SemanticRetrievalMetadata -> MAY_REFERENCE -> SemanticRetrievalProvenance

# ---------- QUERY ----------
rule original_query:
  original SearchQuery MUST remain preserved

rule interpreted_query:
  SemanticQuery interpretation != original query text

rule interpretation_trace:
  semantic mappings SHOULD remain traceable
  to source query spans or context

# ---------- ENTITY / CONCEPT ----------
entity SemanticEntityResolution:
  type: SemanticProcess
  may_include: [mention, candidate_entities, resolved_entity, entity_type, evidence, confidence]

rule entity:
  stable EntityIdentifier SHOULD be preferred over label-only matching

rule ambiguity:
  ambiguous entity mention SHOULD preserve candidates

rule name_identity:
  lexical name equality != entity identity

entity SemanticConceptResolution:
  type: SemanticProcess
  may_include: [query_term, concept_identifier, ontology, preferred_label, mapping_relation, confidence]

rule concept:
  ConceptIdentifier != keyword

rule concept_context:
  concept mapping SHOULD preserve ontology and version

# ---------- RELATIONSHIPS ----------
entity SemanticRelationship:
  type: RelationshipMetadata
  may_include: [subject, predicate, object, relationship_type, ontology, confidence]

rule relationship:
  semantic relationship SHOULD preserve predicate direction

rule cooccurrence:
  lexical co-occurrence != semantic relationship

# ---------- ONTOLOGY / GRAPH ----------
entity RetrievalOntologyContext:
  type: SemanticContext
  may_include: [ontology_identifier, ontology_version, classes, properties, hierarchy, equivalence_relations]

rule ontology:
  ontology version SHOULD be preserved

entity RetrievalKnowledgeGraph:
  type: KnowledgeGraph
  may_include: [graph_identifier, graph_version, entities, relationships, provenance]

rule graph:
  knowledge-graph assertions SHOULD preserve provenance

rule graph_truth:
  graph assertion != independently verified fact

# ---------- SEMANTIC INDEX / EXPANSION ----------
entity SemanticIndex:
  type: SearchIndex
  may_include: [entities, concepts, relationships, semantic_annotations, ontology_mappings, update_time]

rule semantic_index:
  semantic index SHOULD preserve mapping source and version

entity SemanticQueryExpansion:
  type: QueryExpansion
  may_include: [source_term, expanded_term, semantic_relation, ontology, confidence]

semantic_expansion_relations:
  - synonym
  - alias
  - equivalent_concept
  - broader_concept
  - narrower_concept
  - related_concept
  - entity_alias
  - ontology_mapping

rule expansion:
  expansion SHOULD preserve semantic relation to source term

rule expansion_drift:
  broader or related concepts MUST NOT be silently treated as exact equivalents

# ---------- HIERARCHICAL / GRAPH RETRIEVAL ----------
entity HierarchicalRetrieval:
  subtype_of: SemanticRetrievalProcess
  may_include: [start_concept, broader_depth, narrower_depth, hierarchy, traversal_policy]

rule hierarchy:
  broader and narrower relations SHOULD preserve direction

entity SemanticGraphRetrieval:
  subtype_of: SemanticRetrievalProcess
  may_include: [start_entities, predicates, path_constraints, traversal_depth, graph_version]

rule graph_retrieval:
  graph path SHOULD preserve edge types and direction

rule path:
  graph proximity != semantic equivalence

# ---------- VECTOR / HYBRID ----------
entity VectorSemanticRetrieval:
  subtype_of: SemanticRetrievalProcess
  may_include: [query_embedding, vector_index, embedding_model, model_version, metric, top_k]

rule vector:
  query and indexed embeddings MUST occupy compatible embedding space

rule vector_similarity:
  vector similarity != semantic equivalence

entity HybridSemanticRetrieval:
  subtype_of: SemanticRetrievalProcess
  combines:
    - lexical_retrieval_optional
    - entity_retrieval_optional
    - concept_retrieval_optional
    - graph_retrieval_optional
    - vector_retrieval_optional

rule hybrid:
  HybridSemanticRetrieval SHOULD preserve component result sets and fusion method

# ---------- FILTER / SIMILARITY ----------
entity SemanticFilter:
  type: RetrievalFilter
  may_include: [entity_type, concept, relationship, ontology_class, property_constraint]

rule filter:
  semantic filter determines candidate eligibility and != semantic relevance

entity SemanticSimilarity:
  type: SimilarityMetadata
  may_include: [source, target, method, score, model, ontology]

rule similarity:
  semantic similarity SHOULD preserve method and scope

rule similarity_equivalence:
  semantic similarity != semantic equivalence

entity SemanticRelatedness:
  type: SimilarityMetadata
  may_include: [source, target, relationship_basis, score]

rule relatedness:
  semantic relatedness MAY exist without similarity or equivalence

entity SemanticEquivalenceRelation:
  type: SemanticRelationship
  strength: strong

rule equivalence:
  semantic equivalence SHOULD require explicit ontology relation
  or sufficiently strong verified mapping

rule similarity_not_equivalence:
  high similarity MUST NOT alone establish semantic equivalence

# ---------- RELEVANCE / EVIDENCE ----------
entity SemanticRelevance:
  type: RelevanceScore
  may_include: [score, method, matched_entities, matched_concepts, matched_relationships]

rule relevance:
  semantic relevance != factual correctness

rule relevance_confidence:
  relevance != confidence

entity SemanticCandidateEvidence:
  type: RetrievalEvidence
  may_include:
    - lexical_match
    - entity_match
    - concept_match
    - relationship_match
    - ontology_path
    - graph_path
    - embedding_similarity

rule evidence:
  semantic candidate SHOULD preserve retrieval evidence

# ---------- DISAMBIGUATION / MAPPING ----------
entity SemanticDisambiguation:
  type: SemanticProcess
  may_include: [ambiguous_term, candidates, context, selected_candidate, confidence]

rule disambiguation:
  context SHOULD resolve ambiguity without altering original query

rule unresolved:
  unresolved ambiguity SHOULD preserve multiple candidates

entity OntologyMapping:
  type: SemanticRelationship
  may_include:
    - source_concept
    - target_concept
    - mapping_type
    - source_ontology
    - target_ontology
    - confidence

mapping_types:
  - exact
  - close
  - broader
  - narrower
  - related

rule mapping:
  mapping type SHOULD remain explicit

rule close_exact:
  close mapping != exact equivalence

# ---------- TEMPORAL / SPATIAL / RAG ----------
entity SemanticTemporalConstraint:
  type: RetrievalTemporalConstraint
  may_include: [concept_time, valid_time, publication_time, reference_time]

rule temporal:
  temporal semantics SHOULD preserve constrained time dimension

entity SemanticSpatialConstraint:
  type: RetrievalSpatialConstraint
  may_include: [place_entity, geographic_extent, relation, distance]

rule spatial:
  place entity and coordinate geometry SHOULD remain distinguishable

entity SemanticRAGRetrieval:
  subtype_of: SemanticRetrievalProcess
  may_include:
    - semantic_query
    - retrieved_chunks
    - entities
    - concepts
    - retrieval_evidence
    - source_provenance

rule RAG:
  retrieved semantic context SHOULD preserve source and fragment identity

rule RAG_generation:
  retrieved context != generated answer

# ---------- CONFIDENCE / EXPLANATION ----------
entity SemanticRetrievalConfidence:
  type: EpistemicMetadata
  range: [0.0, 1.0]

rule confidence_bounds:
  REQUIRE 0.0 <= confidence <= 1.0

entity SemanticRetrievalExplanation:
  type: RetrievalExplanationMetadata
  may_include:
    - matched_entities
    - matched_concepts
    - matched_relationships
    - ontology_paths
    - graph_paths
    - embedding_scores
    - expansion_relations
    - filters

rule explanation:
  explanation SHOULD expose meaningful semantic retrieval evidence
  without requiring hidden internal chain-of-thought

# ---------- PROVENANCE ----------
entity SemanticRetrievalProvenance:
  type: RetrievalProvenance
  may_include:
    - original_query
    - interpreted_query
    - ontology
    - ontology_version
    - knowledge_graph
    - graph_version
    - semantic_index
    - embedding_model
    - model_version
    - retrieval_strategy
    - timestamp

rule provenance:
  SemanticRetrievalMetadata SHOULD preserve semantic-processing context

# ---------- VALIDATION ----------
validation:
  require:
    - identifiable_semantic_retrieval_process
    - original_query
    - semantic_retrieval_strategy

  reject:
    - confidence_outside_valid_range
    - ambiguous_entity_forced_without evidence
    - keyword_presented_as concept_identifier
    - cooccurrence_presented_as semantic_relationship
    - broader_or_related_concept_presented_as exact_equivalent
    - incompatible_embedding_spaces_compared_directly
    - similarity_presented_as semantic_equivalence_without evidence
    - relevance_presented_as factual_correctness
    - graph_assertion_presented_as verified_fact_without evidence

# ---------- SEMANTIC EQUIVALENCE ----------
SEMANTIC_EQUIVALENCE:
  "semantic retrieval" ~= "meaning-based information retrieval"
  "semantic search retrieval" ~= "semantic retrieval"
  "entity retrieval" ~= "retrieval grounded in entity identity"
  "concept retrieval" ~= "retrieval grounded in ontology concepts"
  "knowledge graph retrieval" ~= "graph-based semantic retrieval"
  "vector semantic retrieval" ~= "embedding-based semantic retrieval"

  PRESERVE distinctions:
    entity != label
    concept != keyword
    relationship != cooccurrence
    semantic_similarity != semantic_equivalence
    semantic_relatedness != semantic_equivalence
    broader_concept != exact_concept
    close_mapping != exact_mapping
    graph_proximity != semantic_equivalence
    relevance != factual_correctness
    confidence != relevance
    retrieved_context != generated_answer

# ---------- CONFIDENCE MODEL ----------
CONFIDENCE:
  explicit_semantic_retrieval_metadata:
    score: 1.00
  stable_entity_identifier:
    score: 0.99
  ontology_grounded_concept:
    score: 0.98
  explicit_semantic_relationship:
    score: 0.98
  verified_exact_ontology_mapping:
    score: 0.98
  compatible_embedding_retrieval:
    score: 0.97
  inferred_entity_resolution:
    score: <= 0.90
  inferred_concept_mapping:
    score: <= 0.85
  lexical_only_semantic_mapping:
    score: <= 0.70
  unresolved_semantic_ambiguity:
    score: <= 0.60
    action: preserve candidates

# ---------- UPDATE MODEL ----------
UPDATE_MODEL:
  on_semantic_query:
    - preserve original query
    - resolve entities and concepts
    - resolve relationships
    - apply ontology and context
    - preserve confidence
    - initialise provenance

  on_semantic_expansion:
    - preserve source query terms
    - apply controlled semantic relations
    - record expansion type
    - prevent unsupported equivalence
    - update provenance

  on_retrieval:
    - query semantic indexes, graphs, or vector indexes
    - preserve component candidate sets
    - preserve semantic evidence
    - apply semantic filters
    - calculate relevance or similarity
    - update provenance

  on_disambiguation:
    - preserve candidate meanings
    - apply context
    - select candidate only when evidence supports resolution
    - preserve confidence and provenance

# ---------- FLOW HASH ----------
FLOW_HASH:
  inputs:
    - ontology
    - semantic_retrieval_entities
    - relationships
    - entity_rules
    - concept_rules
    - relationship_rules
    - ontology_mapping_rules
    - vector_rules
    - validation
    - semantic_equivalence
    - confidence_model
    - update_model

  canonicalisation:
    - preserve_original_query
    - preserve_entity_and_concept_identifiers
    - preserve_ontology_versions
    - preserve_relationship_direction
    - preserve_embedding_model_versions
    - preserve_candidate_evidence
    - preserve_rule_order

  output:
    deterministic_semantic_hash

# ---------- AI CONSUMPTION ----------
AI_CONSUMPTION:
  interpret_as:
    - semantic search metadata ontology
    - meaning-based retrieval model
    - entity, concept, relationship, graph, and vector retrieval structure
    - semantic disambiguation and provenance model

  operations:
    - preserve original query
    - resolve entities
    - map concepts
    - interpret semantic relationships
    - apply ontology context
    - traverse knowledge graphs
    - expand queries through controlled relations
    - perform vector semantic retrieval
    - combine hybrid semantic candidates
    - apply semantic filters
    - disambiguate candidate meanings
    - compare semantic similarity cautiously
    - preserve candidate evidence
    - expose explanation metadata
    - trace provenance
    - preserve uncertainty

  constraints:
    - do_not_equate_label_with entity_identity
    - do_not_equate_keyword_with concept
    - do_not_equate_cooccurrence_with semantic_relationship
    - do_not_equate_similarity_with semantic_equivalence
    - do_not_equate_graph_proximity_with equivalence
    - do_not_present_relevance_as factual_correctness
    - preserve_query_ontology_graph_model_evidence_confidence_and_provenance

# ---------- QUERY MODEL ----------
query:
  input:
    - original_query
    - entity
    - concept
    - relationship
    - ontology
    - knowledge_graph
    - semantic_filter
    - embedding
    - temporal_constraint
    - spatial_constraint

  resolve:
    1. preserve original SearchQuery
    2. resolve entities and candidate identities
    3. resolve concepts against ontology
    4. resolve semantic relationships
    5. apply query context and controlled expansion
    6. query semantic index, graph, or vector index
    7. combine candidate evidence
    8. apply semantic constraints and disambiguation
    9. construct SemanticResultSet and provenance
    10. return confidence

  output:
    - semantic_query
    - resolved_entities
    - resolved_concepts
    - resolved_relationships
    - ontology_context
    - knowledge_graph_context
    - candidates
    - semantic_evidence
    - relevance
    - similarity
    - confidence
    - provenance

# ---------- INFERENCE ----------
inference:
  IF query mention uniquely resolves to stable EntityIdentifier
  THEN attach resolved Entity
  WITH confidence according_to evidence

  IF query term maps to ontology Concept
  THEN attach Concept
  WITH ontology and mapping relation

  IF ontology hierarchy supports broader or narrower traversal
  THEN expand candidates
  WHILE preserving relationship type

  IF compatible embeddings exist
  THEN calculate vector similarity
  WITHOUT inferring semantic equivalence automatically

  IF semantic interpretation remains ambiguous
  THEN preserve candidate entities, concepts, mappings, or paths
  AND reduce confidence

# ---------- OUTPUT CONTRACT ----------
output:
  ontology_entity: SemanticRetrievalMetadata
  semantic_query: source_preserved
  entities: identifier_grounded
  concepts: ontology_grounded
  relationships: direction_preserved
  ontology_and_graph_context: version_aware
  semantic_expansion: relation_preserving
  vector_retrieval: compatibility_aware
  hybrid_retrieval: component_traceable
  semantic_evidence: explicit
  similarity_equivalence_distinction: preserved
  relevance_confidence_distinction: preserved
  disambiguation: confidence_qualified
  RAG_context: source_linked
  provenance: preserved
  uncertainty: explicit
  semantic_retrieval: enabled

end flow