Search Metadata Ontology

FLOW ALGEBRA

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

Authors: Mark Anthony Higgins — Sky (Recursive AI)

flow "SearchMetadataOntology" v1.0

# ---------- DOMAIN ----------
domain:
  class: MetadataOntology
  category: SemanticSearchMetadata
  subject: SearchMetadata
  purpose:
    - represent metadata describing search queries, indexes, retrieval processes, rankings, and results
    - preserve query identity, text, intent, context, filters, retrieval method, relevance, confidence, and provenance
    - support lexical, semantic, vector, structured, and hybrid search
    - distinguish relevance, similarity, confidence, ranking, and factual correctness
    - enable machine-readable reasoning over search processes

# ---------- CORE ENTITIES ----------
entity SearchMetadata:
  type: MetadataEntity
  describes: [SearchQuery, SearchProcess, SearchIndex, SearchResultSet]
  may_include:
    - query_identifier
    - query_text
    - query_language
    - query_type
    - intent
    - entities
    - concepts
    - context
    - filters
    - sorting
    - temporal_constraint
    - spatial_constraint
    - search_index
    - retrieval_method
    - ranking_method
    - relevance
    - similarity
    - confidence
    - result_count
    - latency
    - timestamp
    - provenance

entity SearchQuery:
  type: InformationRequest

entity SearchProcess:
  type: RetrievalProcess
  consumes: SearchQuery
  produces: SearchResultSet

entity SearchIndex:
  type: RetrievalIndex

entity SearchResult:
  type: RetrievalResult

# ---------- RELATIONSHIPS ----------
relationships:
  SearchMetadata -> DESCRIBES -> SearchQuery
  SearchQuery -> MAY_HAVE_INTENT -> SearchIntent
  SearchQuery -> MAY_REFERENCE -> Entity
  SearchQuery -> MAY_REFERENCE -> Concept
  SearchQuery -> MAY_HAVE_CONTEXT -> SearchContext
  SearchQuery -> MAY_HAVE_FILTER -> SearchFilter
  SearchProcess -> USES -> SearchIndex
  SearchProcess -> USES_METHOD -> RetrievalMethod
  SearchProcess -> USES_RANKING -> RankingMethod
  SearchProcess -> RETURNS -> SearchResultSet
  SearchResultSet -> CONTAINS -> SearchResult
  SearchResult -> REFERENCES -> SearchableResource
  SearchResult -> HAS_RANK -> Rank
  SearchResult -> MAY_HAVE_RELEVANCE -> RelevanceScore
  SearchResult -> MAY_HAVE_SIMILARITY -> SimilarityScore
  SearchResult -> MAY_HAVE_CONFIDENCE -> SearchConfidence
  SearchMetadata -> MAY_REFERENCE -> SearchProvenance

# ---------- QUERY ----------
entity QueryIdentifier:
  type: Identifier
  identifies: SearchQuery

rule query_identity:
  SearchQuery SHOULD have QueryIdentifier
  where reproducibility, session tracking, or auditing is required

rule repeated_query:
  identical query text MAY represent distinct executions
  under different time, context, index, or configuration

rule query_text:
  original query text SHOULD be preserved

rule normalised_query:
  normalised query != original query
  and SHOULD remain traceable to source query

rule language:
  query language SHOULD be preserved or inferred with confidence
  where interpretation depends on it

query_types:
  - keyword
  - natural_language
  - question
  - entity_lookup
  - navigational
  - structured
  - semantic
  - vector
  - hybrid

rule query_type:
  QueryType describes retrieval interpretation
  and != SearchIntent

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

rule intent:
  SearchIntent MAY be inferred
  and SHOULD preserve confidence

rule intent_text:
  inferred intent != literal query text

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

rule context:
  SearchContext MAY influence interpretation and ranking
  but MUST NOT alter source facts

# ---------- FILTERS / SORT ----------
entity SearchFilter:
  type: QueryConstraint
  may_include: [field, operator, value, datatype]

filter_operators:
  - equals
  - not_equals
  - greater_than
  - less_than
  - greater_or_equal
  - less_or_equal
  - contains
  - in
  - range
  - exists

rule filter:
  filtering eligibility != ranking relevance

entity SearchSort:
  type: OrderingMetadata
  may_include: [field, direction, priority]

rule sort:
  explicit sort MAY override relevance ordering

# ---------- INDEXES ----------
search_index_types:
  - LexicalIndex
  - SemanticIndex
  - VectorIndex
  - GraphIndex
  - StructuredIndex

rule index:
  SearchIndex SHOULD preserve index identifier, version, or update time
  where reproducibility matters

# ---------- RETRIEVAL ----------
entity LexicalSearch:
  subtype_of: SearchProcess
  may_include:
    - tokenisation
    - stemming
    - lemmatisation
    - term_frequency
    - inverse_document_frequency
    - field_weights

rule lexical:
  lexical match != semantic equivalence

entity SemanticSearch:
  subtype_of: SearchProcess
  may_include:
    - entities
    - concepts
    - ontology
    - semantic_annotations
    - embeddings

rule semantic:
  semantic retrieval SHOULD preserve interpretation method

entity VectorSearch:
  subtype_of: SearchProcess
  may_include:
    - query_embedding
    - vector_index
    - embedding_model
    - model_version
    - distance_metric
    - nearest_neighbour_method

rule vector:
  query and indexed vectors MUST be compatible
  for direct similarity comparison

rule vector_similarity:
  vector similarity != factual correctness

entity HybridSearch:
  subtype_of: SearchProcess
  combines:
    - lexical_score
    - semantic_score
    - vector_score_optional
    - structured_score_optional

rule hybrid:
  HybridSearch SHOULD preserve component scores
  and combination method

retrieval_methods:
  - exact_match
  - boolean
  - full_text
  - BM25
  - semantic_graph
  - vector_nearest_neighbour
  - hybrid
  - filtered_search

entity RetrievalMethod:
  type: SearchMethod

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

expansion_types:
  - synonym
  - alias
  - spelling_variant
  - broader_concept
  - narrower_concept
  - related_concept
  - entity_alias

rule expansion:
  expanded term SHOULD preserve semantic relationship
  to original term

rule expansion_drift:
  unsupported semantic broadening SHOULD NOT be silently applied

# ---------- RELEVANCE / SIMILARITY ----------
entity RelevanceScore:
  type: SearchScore
  may_include: [value, scoring_method, normalisation, scope]

rule relevance:
  RelevanceScore expresses retrieval relevance
  NOT factual truth

rule relevance_comparison:
  scores from incompatible methods SHOULD NOT be directly compared

entity SimilarityScore:
  type: SearchScore
  may_include: [value, metric, representation_space]

rule similarity:
  similarity != semantic equivalence

rule similarity_metric:
  similarity score MUST preserve metric semantics

# ---------- RANKING ----------
entity Rank:
  type: OrdinalMetadata
  range: positive_integer

entity RankingMethod:
  type: SearchMethod
  may_include:
    - algorithm
    - features
    - weights
    - model
    - model_version

rule rank:
  Rank SHOULD preserve ordering convention

rule ranking:
  high rank != factual correctness

entity SearchReranking:
  type: SearchProcess
  may_include:
    - candidate_set
    - reranker
    - reranker_version
    - features
    - output_scores
    - final_order

rule reranking:
  reranking SHOULD preserve original candidate provenance
  and prior scores where auditability matters

# ---------- RESULT SET ----------
entity SearchResultSet:
  type: RetrievalResultCollection
  may_include:
    - result_count
    - returned_count
    - total_estimate
    - results
    - pagination
    - continuation_token

rule result_count:
  returned_count != total_result_count

rule estimated_total:
  estimated total MUST remain distinguishable from exact total

entity SearchResult:
  type: RetrievalResult
  may_include:
    - resource_identifier
    - title
    - URL
    - snippet
    - rank
    - relevance
    - similarity
    - matched_fields
    - matched_entities
    - matched_concepts
    - provenance

rule result:
  SearchResult SHOULD reference source SearchableResource

rule snippet:
  SearchSnippet != full source content

rule generated_snippet:
  generated snippet SHOULD remain distinguishable
  from verbatim extracted snippet

# ---------- PAGINATION ----------
entity SearchPagination:
  type: RetrievalMetadata
  may_include: [page, page_size, offset, cursor, continuation_token]

rule pagination:
  pagination state SHOULD preserve retrieval ordering context

# ---------- TEMPORAL / SPATIAL ----------
entity SearchTemporalConstraint:
  type: TemporalMetadata
  may_include: [start_time, end_time, reference_time, date_field]

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

entity SearchSpatialConstraint:
  type: SpatialMetadata
  may_include: [location, radius, bounding_area, coordinate_reference]

rule spatial:
  spatial search SHOULD preserve coordinate and distance semantics

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

rule confidence_bounds:
  REQUIRE 0.0 <= confidence <= 1.0

rule confidence_scope:
  confidence SHOULD identify whether it qualifies:
    - intent
    - entity_resolution
    - concept_mapping
    - query_expansion
    - result_interpretation

rule confidence_relevance:
  SearchConfidence != RelevanceScore

# ---------- PERFORMANCE ----------
entity SearchPerformanceMetadata:
  type: PerformanceMetadata
  may_include:
    - latency
    - candidate_count
    - index_lookup_time
    - reranking_time

rule performance:
  performance measurements SHOULD preserve search configuration
  where comparison depends on it

# ---------- EXPLANATION ----------
entity SearchExplanationMetadata:
  type: SearchMetadata
  may_include:
    - matched_terms
    - matched_fields
    - matched_entities
    - matched_concepts
    - score_components
    - filters_applied
    - ranking_method

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

# ---------- PROVENANCE ----------
entity SearchProvenance:
  type: ProvenanceMetadata
  may_include:
    - query
    - query_transformations
    - index
    - index_version
    - retrieval_method
    - ranking_method
    - reranker
    - timestamp
    - source_resources

rule provenance:
  SearchMetadata SHOULD preserve retrieval configuration and history

# ---------- VALIDATION ----------
validation:
  require:
    - identifiable_query_or_search_process

  validate_if_present:
    - query_identifier
    - query_text
    - language
    - intent
    - context
    - filters
    - index
    - retrieval_method
    - relevance
    - similarity
    - ranking
    - confidence
    - provenance

  reject:
    - confidence_outside_valid_range
    - incompatible_vector_spaces_compared_without transformation
    - relevance_presented_as factual_confidence
    - similarity_presented_as semantic_equivalence_without evidence
    - rank_presented_as truth
    - inferred_intent_presented_as original_query
    - generated_snippet_presented_as verbatim_source
    - estimated_total_presented_as exact_count

# ---------- SEMANTIC EQUIVALENCE ----------
SEMANTIC_EQUIVALENCE:
  "search metadata" ~= "metadata about search and retrieval"
  "search query metadata" ~= "metadata describing a search query"
  "search result metadata" ~= "metadata describing retrieval results"
  "search ranking metadata" ~= "metadata describing result ordering"
  "query expansion" ~= "search-term expansion" WHEN semantic relationship is preserved
  "hybrid search" ~= "combined lexical and semantic retrieval"

  PRESERVE distinctions:
    query_text != inferred_intent
    query_type != search_intent
    filter != ranking
    lexical_match != semantic_equivalence
    vector_similarity != factual_correctness
    relevance != confidence
    rank != truth
    recency != relevance
    returned_count != total_count
    estimated_total != exact_total
    snippet != source_resource
    explanation_metadata != hidden_chain_of_thought

# ---------- CONFIDENCE MODEL ----------
CONFIDENCE:
  explicit_search_metadata:
    score: 1.00

  explicit_query_and_configuration:
    score: 1.00

  stable_resource_identifier:
    score: 0.99

  compatible_vector_retrieval:
    score: dependent_on_metric

  inferred_query_intent:
    score: <= 0.85

  inferred_entity_resolution:
    score: <= 0.90

  lexical_only_semantic_inference:
    score: <= 0.70

  ambiguous_search_interpretation:
    score: <= 0.60
    action: preserve alternatives

# ---------- UPDATE MODEL ----------
UPDATE_MODEL:
  on_query_creation:
    - assign QueryIdentifier
    - preserve original query text
    - record language
    - initialise context
    - initialise provenance

  on_query_interpretation:
    - preserve original query
    - resolve entities and concepts
    - infer intent with confidence
    - record expansions and transformations
    - update provenance

  on_search_execution:
    - record index and version
    - record retrieval method
    - apply filters and constraints
    - retrieve candidates
    - calculate scores
    - rank results
    - update provenance

  on_reranking:
    - preserve initial result ordering
    - record reranker and version
    - preserve component scores
    - create final ranking
    - update provenance

# ---------- FLOW HASH ----------
FLOW_HASH:
  inputs:
    - ontology
    - search_entities
    - relationships
    - query_rules
    - filter_rules
    - retrieval_rules
    - ranking_rules
    - result_rules
    - validation
    - semantic_equivalence
    - confidence_model
    - update_model

  canonicalisation:
    - normalise_query_metadata
    - preserve_original_query_text
    - preserve_identifiers
    - preserve_result_order
    - preserve_score_method_context
    - preserve_rule_order

  output:
    deterministic_semantic_hash

# ---------- AI CONSUMPTION ----------
AI_CONSUMPTION:
  interpret_as:
    - semantic search metadata ontology
    - query and retrieval lifecycle model
    - ranking and result metadata structure
    - lexical, semantic, vector, and hybrid search model

  operations:
    - identify search queries
    - preserve original query text
    - resolve language and query type
    - infer search intent
    - apply context, filters, and sorting
    - resolve search indexes
    - execute lexical, semantic, vector, or hybrid retrieval
    - expand queries
    - calculate relevance and similarity
    - rank and rerank candidates
    - paginate result sets
    - expose explanation metadata
    - trace provenance
    - preserve uncertainty

  constraints:
    - do_not_equate_query_text_with inferred_intent
    - do_not_equate_filtering_with ranking
    - do_not_equate_similarity_with semantic_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_transformations_indexes_scores_rankings_and_provenance

# ---------- QUERY MODEL ----------
query:
  input:
    - query_text
    - query_type
    - language
    - intent
    - context
    - filter
    - sort
    - temporal_constraint
    - spatial_constraint
    - retrieval_method

  resolve:
    1. create or identify SearchQuery
    2. preserve original query text
    3. resolve language and query type
    4. infer SearchIntent where applicable
    5. apply context, filters, sorting, and constraints
    6. select SearchIndex and RetrievalMethod
    7. retrieve candidate resources
    8. calculate relevance, similarity, and ranking
    9. rerank and construct SearchResultSet
    10. return explanation metadata, provenance, and confidence

  output:
    - search_query
    - interpreted_query
    - intent
    - filters
    - retrieval_method
    - result_set
    - ranks
    - relevance
    - similarity
    - confidence
    - explanation_metadata
    - provenance

# ---------- INFERENCE ----------
inference:
  IF query text exactly matches indexed identifier
  THEN exact-match retrieval MAY receive strong candidate priority

  IF query intent is inferred
  THEN preserve inference confidence
  AND retain original query unchanged

  IF vector query and index share compatible embedding space
  THEN similarity MAY be calculated
  according_to declared metric

  IF lexical and semantic scores are combined
  THEN preserve component scores
  AND apply declared hybrid ranking rule

  IF search interpretation remains ambiguous
  THEN preserve alternative interpretations
  AND reduce confidence

# ---------- OUTPUT CONTRACT ----------
output:
  ontology_entity: SearchMetadata
  query_identity: machine_interpretable
  original_query: preserved
  query_interpretation: explicit
  intent_and_context: confidence_qualified
  filters_and_sorting: explicit
  retrieval_methods: traceable
  indexes: version_aware
  relevance_similarity_confidence: distinguishable
  rankings: ordered_and_traceable
  result_sets: paginatable
  explanation_metadata: available
  provenance: preserved
  uncertainty: explicit
  semantic_retrieval: enabled

end flow