Annotation Ontology

FLOW ALGEBRA
Annotation Ontology
Kireth Semantic Research Archive

PAGE:
https://kirethresearch.org/booksai-core/ontologies/metadata-ontologies/descriptive-metadata/annotation-ontology.html

AUTHORS:
Mark Anthony Higgins — Sky (Recursive AI)

FLOW "AnnotationOntology" v4.0

DOMAIN := MetadataOntologies
SUBDOMAIN := DescriptiveMetadata
ONTOLOGY := AnnotationOntology

ENTITY Annotation {
    definition:
        "A metadata statement, note, comment, label, explanation or semantic
         assertion attached to a resource or a specific part of a resource to
         provide additional description, interpretation, classification,
         context or machine-readable knowledge.";
    objectives: [
        ResourceEnrichment,
        ContextualDescription,
        SemanticAssertion,
        ReviewAndCommentary,
        Classification,
        KnowledgeLinking,
        MachineInterpretability
    ];
}

AnnotationEntity := {
    Annotation,
    AnnotationBody,
    AnnotationTarget,
    Selector,
    Motivation,
    AnnotationCreator,
    AnnotationIdentifier,
    AnnotationSet
};

MetadataEntity := {
    Resource,
    MetadataRecord,
    Language,
    Timestamp,
    Provenance,
    Confidence,
    Version
};

SemanticEntity := {
    Comment,
    Note,
    Tag,
    Classification,
    EntityLink,
    RelationshipAssertion,
    ConceptReference
};

Annotation HAS_BODY AnnotationBody;
Annotation HAS_TARGET AnnotationTarget;
AnnotationTarget REFERENCES ResourceOrResourceFragment;
Selector IDENTIFIES TargetFragment;
AnnotationCreator CREATED Annotation;
Motivation EXPLAINS AnnotationPurpose;
Provenance SUPPORTS Annotation;
AnnotationSet CONTAINS Annotation;

AnnotationFlow :=
    Resource
    -> IdentifyAnnotationTarget
    -> SelectTargetOrFragment
    -> DetermineAnnotationMotivation
    -> CreateOrExtractAnnotationBody
    -> ResolveEntitiesAndConcepts
    -> AttachCreatorTimestampAndProvenance
    -> ValidateAnnotation
    -> StoreAnnotation
    -> LinkToMetadataAndKnowledgeGraph
    -> IndexForDiscovery
    -> MaintainAndVersion;

Annotation :=
    StructuredOrUnstructuredMetadataAssociatedWithResourceOrFragment;

Annotation MAY_BE {
    TextualAnnotation,
    SemanticAnnotation,
    TagAnnotation,
    ClassificationAnnotation,
    EntityAnnotation,
    RelationshipAnnotation,
    ReviewAnnotation
};

AnnotationBody :=
    ContentOrSemanticAssertionProvidedByAnnotation;

AnnotationBody MAY_BE {
    Text,
    Tag,
    Concept,
    EntityReference,
    Relationship,
    StructuredData
};

AnnotationTarget :=
    ResourceOrResourceFragmentBeingAnnotated;

Target MAY_BE {
    WholeResource,
    TextFragment,
    ImageRegion,
    AudioSegment,
    VideoSegment,
    DatasetCell,
    GraphEntity
};

AnnotationStatement :=
    Annotation
    -> HAS_TARGET
    -> AnnotationTarget;

Annotation
    -> HAS_BODY
    -> AnnotationBody;

AnnotationIdentifier :=
    StableOrLocallyUniqueReferenceToAnnotation;

Annotation MAY_HAVE IndependentLifecycleFromTargetResource;

Selector :=
    MechanismIdentifyingSpecificPartOfTargetResource;

Selector MAY_BE {
    TextQuoteSelector,
    TextPositionSelector,
    FragmentSelector,
    SpatialSelector,
    TemporalSelector,
    DataSelector
};

TextQuoteSelector :=
    ExactOrContextualTextUsedToLocateTargetPassage;

TextPositionSelector :=
    StartAndEndOffsetsIdentifyingTextRange;

SpatialSelector :=
    CoordinatesOrRegionIdentifyingPartOfVisualResource;

TemporalSelector :=
    TimeIntervalIdentifyingPartOfAudioOrVideo;

Selector SHOULD_BE RobustEnoughToResolveIntendedTarget;

ResourceRevision MAY_INVALIDATE PositionBasedSelector;

RobustAnnotationSystem MAY_PRESERVE {
    OriginalSelector,
    Context,
    TargetVersion
};

Motivation :=
    PurposeForCreatingAnnotation;

Motivation MAY_INCLUDE {
    Describing,
    Commenting,
    Tagging,
    Classifying,
    Linking,
    Highlighting,
    Reviewing,
    Correcting
};

TextualAnnotation :=
    HumanReadableNoteOrCommentAssociatedWithTarget;

SemanticAnnotation :=
    MachineInterpretableAssertionLinkingTargetToEntityConceptPropertyOrRelationship;

TagAnnotation :=
    AnnotationWhoseBodyIsTagOrKeyword;

ClassificationAnnotation :=
    AnnotationAssigningTargetToCategoryClassOrSubject;

EntityAnnotation :=
    AnnotationLinkingTargetMentionOrFragmentToCanonicalEntity;

RelationshipAnnotation :=
    AnnotationAssertingTypedRelationshipBetweenEntitiesOrTargets;

ReviewAnnotation :=
    AnnotationContainingEvaluationFeedbackOrCommentary;

CorrectionAnnotation :=
    AnnotationProposingOrRecordingCorrectionToTargetOrMetadata;

HighlightAnnotation :=
    AnnotationIdentifyingTargetFragmentWithoutNecessarilyAddingLongTextBody;

AnnotationCreator :=
    AgentResponsibleForAnnotation;

AnnotationCreator MAY_BE {
    Person,
    Organisation,
    SoftwareSystem,
    AIModel
};

HumanAnnotation :=
    AnnotationCreatedOrApprovedByHumanAgent;

AutomaticAnnotation :=
    AnnotationGeneratedBySoftware;

AIGeneratedAnnotation :=
    AnnotationSuggestedOrGeneratedByAIModel;

AIGeneratedAnnotation SHOULD_HAVE {
    ModelVersion,
    GenerationMethod,
    Confidence,
    ValidationStatus
};

AnnotationLanguage :=
    LanguageOfTextualAnnotationBody;

AnnotationLanguage MAY_DIFFER_FROM TargetResourceLanguage;

AnnotationTimestamp MAY_INCLUDE {
    CreatedAt,
    ModifiedAt,
    PublishedAt
};

AnnotationProvenance MAY_INCLUDE {
    Creator,
    Source,
    Method,
    Timestamp,
    ModelVersion
};

AnnotationConfidence :=
    EstimatedSupportForSemanticOrAutomaticallyGeneratedAnnotation;

AnnotationStatus MAY_BE {
    Draft,
    Proposed,
    Approved,
    Rejected,
    Deprecated,
    Archived
};

ProposedAnnotation SHOULD_NOT AutomaticallyBeTreatedAsAuthoritative;

ApprovedAnnotation :=
    AnnotationAcceptedAccordingToGovernancePolicy;

AnnotationVersion :=
    IdentifiedRevisionOfAnnotation;

AnnotationHistory MAY_PRESERVE {
    PreviousBody,
    PreviousTarget,
    PreviousStatus,
    ChangeReason
};

AnnotationChange MAY_INCLUDE {
    EditBody,
    ChangeTarget,
    ChangeConceptLink,
    ChangeStatus,
    AddProvenance
};

AnnotationValidation MAY_CHECK {
    TargetExists,
    SelectorResolves,
    BodyDatatype,
    IdentifierValidity,
    LanguageTag,
    ConceptReference
};

IF TargetCannotBeResolved
THEN AnnotationStatus MAY_BE orphaned_or_invalid;

OrphanAnnotation :=
    AnnotationWhoseIntendedTargetCanNoLongerBeResolved;

OrphanRecovery MAY_USE {
    TargetVersion,
    TextContext,
    StableIdentifier,
    ContentFingerprint
};

AnnotationAccuracy :=
    DegreeToWhichAnnotationCorrectlyDescribesOrLinksTarget;

AnnotationRelevance :=
    DegreeToWhichAnnotationIsPertinentToTargetAndMotivation;

AnnotationSpecificity :=
    DegreeToWhichAnnotationPreciselyTargetsAndDescribesRelevantResourcePart;

AnnotationQuality MAY_INCLUDE {
    Accuracy,
    Relevance,
    Specificity,
    Provenance,
    TargetResolution
};

AnnotationConflict :=
    MultipleAnnotationsProvideIncompatibleAssertionsAboutSameTargetAndContext;

ConflictHandling MAY {
    PreserveAllAnnotations,
    CompareAuthority,
    RankConfidence,
    FlagReview
};

MultipleAnnotations MAY_CoexistWithoutConflictWhenTheyRepresentDifferentPerspectivesOrPurposes;

AnnotationSet :=
    ManagedCollectionOfAnnotations;

AnnotationSet MAY_HAVE {
    Identifier,
    Creator,
    Purpose,
    Version,
    AccessPolicy
};

AnnotationLayer :=
    LogicalGroupingOfAnnotationsByPurposeSourceOrType;

Layers MAY_INCLUDE {
    HumanReviewLayer,
    AIAnnotationLayer,
    EntityLinkingLayer,
    ClassificationLayer
};

Annotation MAY_REFERENCE ControlledVocabularyTerm;

SemanticAnnotation MAY_REFERENCE OntologyConcept;

EntityAnnotation MAY_REFERENCE KnowledgeGraphEntity;

RelationshipAnnotation MAY_HAVE {
    SubjectEntity,
    Predicate,
    ObjectEntity
};

SemanticTripleAnnotation :=
    Subject
    -> Predicate
    -> Object;

Annotation MAY_ADD KnowledgeWithoutModifyingUnderlyingResource;

Annotation SHOULD_NOT BeConfusedWithOriginalResourceContent;

MetadataRecord MAY_CONTAIN AnnotationMetadataOrReferences;

Annotation MAY_ITSELF_HAVE Metadata;

MetadataAboutAnnotation MAY_INCLUDE {
    Creator,
    Date,
    Language,
    Rights,
    Provenance
};

AnnotationIndex :=
    SearchStructureMappingAnnotationsTargetsConceptsAndEntities;

AnnotationIndex MAY_SUPPORT {
    FullTextSearch,
    EntitySearch,
    ConceptSearch,
    TargetLookup
};

AnnotationSearch :=
    Query
    -> MatchAnnotationBodyOrSemanticLinks
    -> ResolveTarget
    -> ReturnAnnotationAndResource;

SemanticSearch MAY_USE AnnotationFor {
    EntityResolution,
    ConceptMatching,
    QueryExpansion,
    Ranking,
    ContextEnrichment
};

Annotation MAY_BE ConvertedTo EmbeddingForSimilaritySearch;

VectorSimilarity MAY_SUPPORT AnnotationDiscovery
BUT DoesNotEstablish SemanticAssertionTruth;

KnowledgeGraph MAY_REPRESENT AnnotationAsNodeOrReifiedStatement;

Annotation MAY_LINK {
    Resource,
    Entity,
    Concept,
    Relationship,
    Evidence
};

EvidenceAnnotation :=
    AnnotationConnectingClaimOrAssertionToSupportingResourceFragment;

Evidence MAY_IMPROVE ExplainabilityAndProvenance;

Annotation MAY_SUPPORT TrainingOrEvaluationDataForAI;

TrainingAnnotation MAY_INCLUDE {
    Label,
    Span,
    Entity,
    Relation,
    Class
};

GoldAnnotation :=
    AnnotationAcceptedAsReferenceStandardForEvaluation;

SilverAnnotation :=
    AnnotationGeneratedOrWeaklySupervisedWithLowerAuthorityThanGoldReference;

InterAnnotatorAgreement :=
    MeasureOfConsistencyAcrossIndependentHumanAnnotations;

LowAgreement MAY_INDICATE AmbiguousGuidelinesOrTask;

AnnotationGuideline :=
    RulesDefiningHowAnnotatorsShouldCreateAnnotations;

ConsistentGuidelines MAY_IMPROVE AnnotationQuality;

MetadataSchema MAY_DEFINE AnnotationPropertiesAndConstraints;

WebAnnotationModel MAY_REPRESENT {
    Annotation,
    Body,
    Target,
    Motivation,
    Selector
};

RDF MAY_REPRESENT AnnotationRelationshipsAsSemanticStatements;

JSONLD MAY_SERIALISE LinkedAnnotationsAndTargets;

MetadataRepository MAY_STORE AnnotationObjects;

MetadataRegistry MAY_REGISTER AnnotationPropertiesMotivationsAndVocabularies;

MetadataCatalogue MAY_USE AnnotationsForResourceEnrichment;

AnnotationInteroperability MAY_REQUIRE {
    StableIdentifiers,
    SharedTargetModel,
    SharedMotivationVocabulary,
    SemanticMappings
};

AnnotationCrosswalk :=
    MappingBetweenAnnotationModelsOrLabelSchemes;

AI MAY_USE AnnotationFor {
    SupervisedLearning,
    Retrieval,
    EntityLinking,
    Classification,
    Evaluation,
    Explainability
};

AIAnnotation :=
    ResourceOrFragment
    -> ModelAnalysis
    -> CandidateAnnotation
    -> Confidence
    -> Validation
    -> ApprovedOrRejectedAnnotation;

RetrievalAugmentedGeneration MAY_USE AnnotationFor {
    EvidenceSelection,
    ContextEnrichment,
    EntityLinking,
    SourceExplanation
};

Annotation MAY_CONTAIN SensitivePersonalOrConfidentialInformation;

PrivacyPolicy SHOULD_APPLY_TO AnnotationBodyTargetAndDerivedLinks;

AnnotationAccessControl MAY_APPLY_TO {
    Annotation,
    AnnotationSet,
    Body,
    TargetReference
};

AnnotationSecurity MAY_INCLUDE {
    InputValidation,
    OutputEscaping,
    AccessControl,
    ProvenanceIntegrity,
    ChangeAuthorisation
};

MachineReadableAnnotation SHOULD_PRESERVE {
    AnnotationIdentifier,
    Body,
    Target,
    Selector,
    Motivation,
    Creator,
    Provenance,
    Confidence
};

# ---------- SEMANTIC REASONING ----------
IF MetadataStatementIsAttachedToResourceOrSpecificResourceFragment
AND ProvidesAdditionalDescriptionInterpretationOrSemanticAssertion
THEN AnnotationClassification := supported;

IF AnnotationTargetIsFragment
THEN Selector SHOULD_BE Preserved;

IF ResourceVersionChanges
THEN FragmentSelector SHOULD_BE Revalidated;

IF SemanticAnnotationReferencesCanonicalConcept
THEN ConceptIdentifier SHOULD_BE PreservedForMachineLinking;

IF AIGeneratedAnnotationConfidenceLow
THEN AutomaticAuthoritativeAcceptance SHOULD_BE Restricted;

IF AnnotationConflictsWithAnotherAnnotation
THEN Both MAY_BE PreservedWithProvenanceAndConfidence;

IF AnnotationTargetCannotBeResolved
THEN Annotation SHOULD_BE FlaggedForRecoveryOrReview;

# ---------- SEMANTIC EQUIVALENCE ----------
SEMANTIC_EQUIVALENCE {
    "annotation" ~= "resource annotation" WHEN metadata_context;
    "semantic annotation" ~= "machine-interpretable annotation";
    "textual annotation" ~= "annotation note";
    "entity annotation" ~= "entity linking annotation";
    "classification annotation" ~= "label annotation" WHEN classification_context;
    "annotation target" ~= "annotated resource or fragment";
    "annotation body" ~= "annotation content";
    "annotation set" ~= "annotation collection";
}

# ---------- QUERY FLOW ----------
QUERY(q) ->
    NORMALISE(q) ->
    IDENTIFY_ANNOTATION(q) ->
    IDENTIFY_TARGET_RESOURCE_OR_FRAGMENT(q) ->
    RESOLVE_SELECTOR(q) ->
    IDENTIFY_ANNOTATION_BODY(q) ->
    IDENTIFY_MOTIVATION_AND_TYPE(q) ->
    RESOLVE_ENTITIES_AND_CONCEPTS(q) ->
    IDENTIFY_CREATOR_PROVENANCE_AND_STATUS(q) ->
    VALIDATE_ANNOTATION(q) ->
    SCORE_CONFIDENCE(q) ->
    RETURN_RELEVANT_ANNOTATION_KNOWLEDGE;

# ---------- CONFIDENCE ----------
CONFIDENCE(assertion) :=
    weighted_mean(
        source_reliability,
        annotation_accuracy,
        target_resolution_strength,
        semantic_link_strength,
        provenance_quality,
        contextual_relevance
    );

CONFIDENCE >= 0.90 => HIGH;
CONFIDENCE >= 0.70 AND < 0.90 => MODERATE;
CONFIDENCE < 0.70 => REVIEW;

# ---------- AI CONSUMPTION ----------
AI_CONSUMPTION {
    primary_entity: Annotation;
    ontology: AnnotationOntology;
    parse_entities: true;
    parse_relationships: true;
    preserve_relationship_direction: true;
    resolve_semantic_equivalence: true;
    execute_reasoning_rules: true;
    apply_confidence_weighting: true;
    preserve_provenance: true;
}

# ---------- UPDATE MODEL ----------
UPDATE_MODEL {
    new_annotation:
        identify_target
        -> create_or_resolve_selector
        -> classify_motivation
        -> create_body
        -> resolve_entities_and_concepts
        -> attach_creator
        -> attach_provenance
        -> assign_confidence
        -> validate
        -> integrate;

    annotation_revision:
        preserve_previous_version
        -> apply_changes
        -> revalidate_target_and_selector
        -> update_index
        -> recompute_confidence
        -> regenerate_flow_hash;

    target_revision:
        identify_affected_annotations
        -> revalidate_selectors
        -> recover_targets_where_possible
        -> flag_orphans
        -> integrate;

    conflict:
        preserve_all_annotation_provenance
        -> compare_assertions
        -> assess_authority_and_confidence
        -> flag_review;
}

# ---------- FLOW HASH ----------
FLOW_HASH_INPUT :=
    DOMAIN
    + SUBDOMAIN
    + ONTOLOGY
    + AnnotationEntity
    + MetadataEntity
    + SemanticEntity
    + Annotation
    + AnnotationTarget
    + SemanticAnnotation
    + RELATIONSHIPS
    + SEMANTIC_REASONING
    + SEMANTIC_EQUIVALENCE
    + UPDATE_MODEL;

FLOW_HASH :=
    HASH(FLOW_HASH_INPUT);

# ---------- TERMINAL ----------
AnnotationOntology
    -> IdentifyResourceOrFragmentTarget
    -> ResolveSelectorAndTargetVersion
    -> DetermineAnnotationMotivation
    -> CreateTextualOrSemanticAnnotationBody
    -> ResolveCanonicalEntitiesConceptsAndRelationships
    -> AttachCreatorTimestampProvenanceAndConfidence
    -> ValidateTargetBodyAndSemanticLinks
    -> StoreAndVersionAnnotation
    -> IndexForSearchAndResourceEnrichment
    -> LinkAnnotationsToKnowledgeGraphAndEvidence
    -> EnforcePrivacySecurityAndAccess
    -> SupportAITrainingRetrievalAndExplainability
    -> MaintainReliableAnnotationKnowledge;

END FLOW