Semantic Search Metadata

FLOW ALGEBRA

Ontology: Semantic Search Metadata Ontology
Domain: Metadata Ontologies > Semantic Search Metadata
Canonical URL: https://kirethresearch.org/booksai-core/ontologies/metadata-ontologies/semantic-search-metadata.html

Authors: Mark Anthony Higgins — Sky (Recursive AI)

flow "SemanticSearchMetadataOntology" v1.0

# ---------- DOMAIN ----------
domain:
  class: MetadataOntology
  category: SemanticSearchMetadata
  subject: SemanticSearchMetadata
  purpose:
    - represent metadata supporting semantic search, retrieval, ranking, and machine interpretation
    - preserve entities, concepts, relationships, intent, context, embeddings, relevance, confidence, and provenance
    - support meaning-based discovery beyond lexical keyword matching
    - connect queries, indexed resources, ontologies, knowledge graphs, vector representations, and retrieval results
    - enable machine-readable reasoning over semantic search processes

# ---------- CORE ENTITY ----------
entity SemanticSearchMetadata:
  type: MetadataEntity
  describes:
    - SearchableResource
    - SemanticQuery
    - RetrievalProcess
    - SearchResult
  may_include:
    - resource_identifier
    - entity_metadata
    - concept_metadata
    - relationship_metadata
    - ontology_reference
    - knowledge_graph_reference
    - semantic_annotations
    - embedding_metadata
    - query_intent
    - query_context
    - relevance_metadata
    - ranking_metadata
    - confidence
    - provenance
    - temporal_metadata
    - spatial_metadata

entity SearchableResource:
  type: InformationResource

entity SemanticQuery:
  type: SearchQuery
  seeks: MeaningfulResource

entity SemanticSearchProcess:
  type: RetrievalProcess
  consumes: SemanticQuery
  produces: SearchResultSet

# ---------- RELATIONSHIPS ----------
relationships:
  SemanticSearchMetadata -> DESCRIBES -> SearchableResource
  SemanticSearchMetadata -> MAY_DESCRIBE -> SemanticQuery
  SearchableResource -> MAY_MENTION -> Entity
  SearchableResource -> MAY_EXPRESS -> Concept
  SearchableResource -> MAY_ASSERT -> Relationship
  SearchableResource -> MAY_REFERENCE_ONTOLOGY -> Ontology
  SearchableResource -> MAY_HAVE_EMBEDDING -> Embedding
  SemanticQuery -> MAY_REFERENCE -> Entity
  SemanticQuery -> MAY_EXPRESS -> SearchIntent
  SemanticQuery -> HAS_CONTEXT -> SearchContext
  SemanticSearchProcess -> USES -> SemanticIndex
  SemanticSearchProcess -> MAY_USE -> KnowledgeGraph
  SemanticSearchProcess -> MAY_USE -> VectorIndex
  SemanticSearchProcess -> RETURNS -> SearchResult
  SearchResult -> HAS_RELEVANCE -> RelevanceScore
  SearchResult -> MAY_HAVE_CONFIDENCE -> SearchConfidence
  SemanticSearchMetadata -> MAY_REFERENCE -> SearchProvenance

# ---------- RESOURCE IDENTITY ----------
rule resource_identity:
  SearchableResource SHOULD have stable identifier
  where persistent indexing and retrieval are required

rule label_not_identity:
  resource label or title MUST NOT be assumed unique identity

# ---------- ENTITY METADATA ----------
entity SearchEntityMetadata:
  type: EntityMetadata
  may_include:
    - entity_identifier
    - entity_type
    - label
    - aliases
    - relationships
    - confidence

rule entity:
  semantic search SHOULD prefer stable entity identifiers
  over lexical label matching where available

rule entity_resolution:
  ambiguous mentions SHOULD preserve candidate entities
  until sufficient context resolves identity

# ---------- CONCEPT METADATA ----------
entity SearchConceptMetadata:
  type: ConceptMetadata
  may_include:
    - concept_identifier
    - preferred_label
    - alternative_labels
    - broader_concept
    - narrower_concept
    - related_concept
    - ontology

rule concept:
  ConceptIdentifier SHOULD remain distinct
  from keyword or surface phrase

# ---------- RELATIONSHIPS ----------
entity SearchRelationshipMetadata:
  type: RelationshipMetadata
  may_include:
    - subject
    - predicate
    - object
    - relationship_type
    - confidence
    - provenance

rule relationship:
  semantic relationship SHOULD preserve direction and predicate semantics

rule lexical_cooccurrence:
  co-occurrence != semantic relationship
  unless relationship evidence supports it

# ---------- ONTOLOGY CONTEXT ----------
entity SearchOntologyContext:
  type: SemanticContext
  may_include:
    - ontology_identifier
    - ontology_version
    - classes
    - properties
    - equivalence_rules
    - hierarchy

rule ontology:
  ontology context SHOULD preserve version
  where concept or relationship semantics may change

# ---------- KNOWLEDGE GRAPH ----------
entity SearchKnowledgeGraph:
  type: KnowledgeGraph
  may_include:
    - graph_identifier
    - entities
    - relationships
    - provenance
    - update_time

rule knowledge_graph:
  graph assertions SHOULD remain distinguishable
  from independently verified facts
  unless verification metadata supports them

# ---------- SEMANTIC ANNOTATIONS ----------
entity SemanticAnnotation:
  type: SemanticMetadata
  may_include:
    - target_resource
    - entity
    - concept
    - relationship
    - span
    - annotation_method
    - confidence

rule annotation:
  machine-generated annotation SHOULD preserve method and confidence

rule annotation_not_truth:
  SemanticAnnotation != verified semantic fact

# ---------- QUERY ----------
entity SemanticQueryMetadata:
  type: SearchMetadata
  may_include:
    - query_identifier
    - query_text
    - language
    - entities
    - concepts
    - intent
    - context
    - filters
    - temporal_constraint
    - spatial_constraint
    - embedding

rule query:
  original query SHOULD be preserved
  alongside semantic interpretation

rule interpretation:
  query interpretation MUST remain distinguishable
  from literal query text

# ---------- INTENT ----------
entity SearchIntent:
  type: SemanticMetadata
  may_be:
    - informational
    - navigational
    - transactional
    - comparative
    - local
    - research
    - entity_lookup
    - question_answering

rule intent:
  SearchIntent MAY be inferred
  and SHOULD preserve confidence

rule intent_not_query:
  inferred intent != original query text

# ---------- CONTEXT ----------
entity SearchContext:
  type: ContextMetadata
  may_include:
    - conversation_context
    - user_context
    - domain_context
    - temporal_context
    - spatial_context
    - task_context

rule context:
  search context MAY alter relevance interpretation
  but MUST NOT alter source facts

# ---------- QUERY EXPANSION ----------
entity QueryExpansion:
  type: SearchProcess
  may_include:
    - original_term
    - expanded_term
    - relation
    - source
    - confidence

expansion_relations:
  - synonym
  - alias
  - broader
  - narrower
  - related
  - entity_alias
  - spelling_variant

rule expansion:
  expanded terms SHOULD preserve semantic relationship
  to original query

rule expansion_drift:
  query expansion MUST NOT silently broaden meaning
  beyond supported semantic relation

# ---------- EMBEDDINGS ----------
entity SearchEmbeddingMetadata:
  type: VectorMetadata
  may_include:
    - embedding_identifier
    - embedding_model
    - model_version
    - vector_dimension
    - distance_metric
    - normalisation
    - creation_time

rule embedding:
  embedding SHOULD preserve generating model and version

rule embedding_space:
  embeddings from incompatible models or spaces
  MUST NOT be assumed directly comparable

# ---------- VECTOR INDEX ----------
entity VectorIndex:
  type: SemanticIndex
  may_include:
    - index_identifier
    - embedding_model
    - vector_dimension
    - distance_metric
    - indexing_method
    - update_time

rule vector_index:
  query embedding and indexed embeddings SHOULD be compatible

# ---------- LEXICAL INDEX ----------
entity LexicalIndex:
  type: SearchIndex
  may_include:
    - tokenisation
    - stemming
    - language
    - ranking_method

# ---------- HYBRID SEARCH ----------
entity HybridSearch:
  subtype_of: SemanticSearchProcess
  combines:
    - lexical_retrieval
    - semantic_retrieval

rule hybrid:
  hybrid ranking SHOULD preserve component scores
  and combination method where available

# ---------- RELEVANCE ----------
entity RelevanceScore:
  type: SearchScore
  may_include:
    - value
    - method
    - scope
    - normalisation

rule relevance:
  relevance score != factual confidence

rule score_comparison:
  scores from different ranking methods
  MUST NOT be assumed directly comparable

# ---------- SIMILARITY ----------
entity SimilarityScore:
  type: SearchScore
  may_include:
    - value
    - metric
    - embedding_space

rule similarity:
  semantic similarity != semantic equivalence

rule distance:
  similarity and distance MUST preserve metric semantics

# ---------- RANKING ----------
entity RankingMetadata:
  type: SearchMetadata
  may_include:
    - rank
    - score
    - ranking_method
    - component_scores
    - reranker
    - tie_breaker

rule rank:
  lower numerical rank MAY indicate higher result position
  according_to ranking convention

rule ranking_not_truth:
  high rank != factual correctness

# ---------- RERANKING ----------
entity SemanticReranking:
  type: SearchProcess
  may_include:
    - candidate_set
    - reranker_model
    - features
    - output_scores
    - final_rank

rule reranking:
  reranked results SHOULD preserve original candidate provenance
  where auditability is required

# ---------- FILTERS ----------
entity SearchFilter:
  type: QueryConstraint
  may_include:
    - field
    - operator
    - value

rule filter:
  SearchFilter SHOULD be applied separately
  from semantic similarity unless ranking method explicitly combines them

# ---------- TEMPORAL SEARCH ----------
rule temporal_search:
  temporal constraints SHOULD preserve:
    - reference_time
    - valid_from
    - valid_until
    - publication_time
    - modification_time

rule recency:
  recency != relevance
  unless ranking method explicitly incorporates recency

# ---------- SPATIAL SEARCH ----------
rule spatial_search:
  spatial constraints SHOULD preserve:
    - location
    - distance
    - geographic_extent
    - coordinate_reference

# ---------- RESULT ----------
entity SearchResult:
  type: RetrievalResult
  may_include:
    - resource
    - rank
    - relevance
    - similarity
    - matched_entities
    - matched_concepts
    - explanation_metadata
    - provenance

rule result:
  SearchResult SHOULD link to source SearchableResource

rule snippet:
  generated or extracted snippet != complete source resource

# ---------- EXPLANATION METADATA ----------
entity SearchExplanationMetadata:
  type: SearchMetadata
  may_include:
    - matched_terms
    - matched_entities
    - matched_concepts
    - semantic_paths
    - score_components
    - filters_applied

rule explanation:
  explanation metadata SHOULD describe externally meaningful ranking evidence
  without requiring hidden internal chain-of-thought

# ---------- CONFIDENCE ----------
entity SearchConfidence:
  type: EpistemicMetadata
  range: [0.0, 1.0]

rule confidence_bounds:
  REQUIRE 0.0 <= confidence <= 1.0

rule confidence:
  SearchConfidence SHOULD identify what it qualifies:
    - entity_resolution
    - intent_detection
    - semantic_annotation
    - relationship_mapping
    - result_interpretation

rule confidence_relevance:
  confidence != relevance

# ---------- PROVENANCE ----------
entity SearchProvenance:
  type: ProvenanceMetadata
  may_include:
    - source_resource
    - index
    - embedding_model
    - ontology
    - knowledge_graph
    - retrieval_method
    - ranking_method
    - reranker
    - retrieval_time

rule provenance:
  SemanticSearchMetadata SHOULD preserve retrieval and semantic-processing context

# ---------- VALIDATION ----------
validation:
  require:
    - identifiable_searchable_resource_or_query

  validate_if_present:
    - entities
    - concepts
    - relationships
    - ontology
    - embeddings
    - query_intent
    - context
    - relevance
    - similarity
    - ranking
    - confidence
    - provenance

  reject:
    - confidence_outside_valid_range
    - incompatible_embedding_spaces_compared_without transformation
    - relevance_presented_as factual_confidence
    - similarity_presented_as semantic_equivalence_without evidence
    - ranking_presented_as truth
    - inferred_intent_presented_as literal_query
    - query_expansion_without semantic_relationship

# ---------- SEMANTIC EQUIVALENCE ----------
SEMANTIC_EQUIVALENCE:
  "semantic search metadata" ~= "metadata supporting meaning-based search"
  "semantic retrieval metadata" ~= "semantic search metadata"
  "query intent" ~= "search intent"
  "semantic annotation" ~= "entity or concept annotation" WHEN semantic targets are explicit
  "vector search metadata" ~= "embedding-based retrieval metadata"
  "hybrid search" ~= "combined lexical and semantic retrieval"

  PRESERVE distinctions:
    entity != label
    concept != keyword
    relationship != cooccurrence
    query_text != inferred_intent
    embedding_similarity != semantic_equivalence
    relevance != confidence
    rank != truth
    recency != relevance
    annotation != verified_fact
    snippet != source_resource
    explanation_metadata != hidden_chain_of_thought

# ---------- CONFIDENCE MODEL ----------
CONFIDENCE:
  explicit_semantic_metadata:
    score: 1.00

  stable_entity_identifier:
    score: 0.99

  ontology_grounded_concept:
    score: 0.98

  explicit_relationship:
    score: 0.98

  compatible_embedding_match:
    score: dependent_on_metric

  inferred_entity_resolution:
    score: <= 0.90

  inferred_query_intent:
    score: <= 0.85

  lexical_only_semantic_inference:
    score: <= 0.70

  ambiguous_semantic_mapping:
    score: <= 0.60
    action: preserve candidates

# ---------- UPDATE MODEL ----------
UPDATE_MODEL:
  on_resource_index:
    - assign or resolve resource identifier
    - extract entities and concepts
    - attach semantic annotations
    - generate embeddings where applicable
    - update semantic index
    - initialise provenance

  on_query:
    - preserve original query
    - resolve entities and concepts
    - infer intent with confidence
    - apply context and filters
    - generate query embedding where applicable
    - initialise retrieval provenance

  on_retrieval:
    - retrieve lexical and/or semantic candidates
    - preserve component scores
    - apply filters
    - rank candidates
    - preserve provenance

  on_reranking:
    - preserve original candidate order and scores
    - apply reranker
    - record final rank
    - update explanation metadata and provenance

# ---------- FLOW HASH ----------
FLOW_HASH:
  inputs:
    - ontology
    - semantic_search_entities
    - relationships
    - entity_rules
    - concept_rules
    - query_rules
    - embedding_rules
    - ranking_rules
    - validation
    - semantic_equivalence
    - confidence_model
    - update_model

  canonicalisation:
    - normalise_whitespace
    - preserve_semantic_identifiers
    - preserve_embedding_model_version
    - preserve_ranking_order
    - sort_unordered_annotation_sets
    - preserve_rule_order

  output:
    deterministic_semantic_hash

# ---------- AI CONSUMPTION ----------
AI_CONSUMPTION:
  interpret_as:
    - metadata ontology
    - semantic search and retrieval model
    - entity, concept, and relationship indexing structure
    - vector, lexical, and hybrid ranking model

  operations:
    - identify searchable resources
    - resolve entities and concepts
    - traverse semantic relationships
    - apply ontology context
    - interpret query intent
    - apply search context
    - expand queries semantically
    - generate and compare compatible embeddings
    - retrieve lexical and semantic candidates
    - combine hybrid scores
    - rerank results
    - apply temporal and spatial filters
    - explain ranking metadata
    - trace search provenance
    - preserve uncertainty

  constraints:
    - do_not_equate_label_with entity_identity
    - do_not_equate_keyword_with concept
    - do_not_equate_similarity_with equivalence
    - do_not_equate_relevance_with factual_confidence
    - do_not_equate_rank_with truth
    - do_not_require_or_expose_hidden_chain_of_thought
    - preserve_query_context_models_scores_rankings_and_provenance

# ---------- QUERY MODEL ----------
query:
  input:
    - query_text
    - entity
    - concept
    - relationship
    - intent
    - domain_context
    - temporal_constraint
    - spatial_constraint
    - filter
    - retrieval_mode

  resolve:
    1. preserve original SemanticQuery
    2. resolve entities and concepts
    3. infer SearchIntent
    4. apply ontology and context
    5. perform semantic query expansion
    6. generate compatible query embedding where applicable
    7. retrieve candidates from semantic and lexical indexes
    8. apply filters and reranking
    9. construct explanation metadata and provenance
    10. return ranked SearchResultSet

  output:
    - interpreted_query
    - entities
    - concepts
    - intent
    - expanded_terms
    - retrieval_method
    - results
    - ranks
    - relevance
    - confidence
    - provenance

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

  IF query term maps to ontology Concept
  THEN expand with controlled semantic relations
  subject_to relation type

  IF query and resource embeddings share compatible model space
  THEN calculate semantic similarity
  according_to declared metric

  IF lexical and semantic retrieval are combined
  THEN preserve both component scores
  AND apply declared hybrid ranking method

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

# ---------- OUTPUT CONTRACT ----------
output:
  ontology_entity: SemanticSearchMetadata
  searchable_resources: semantically_annotated
  query_interpretation: explicit
  entities_and_concepts: identifier_grounded
  relationships: traversable
  embeddings: model_version_aware
  lexical_semantic_hybrid_retrieval: supported
  relevance_and_confidence: distinguishable
  rankings: traceable
  explanation_metadata: available
  provenance: preserved
  uncertainty: explicit
  semantic_retrieval: enabled

end flow