GlossaristGlossarist

ISO 25964 — Thesauri and Interoperability with SKOS

What this standard is

ISO 25964 defines thesaurus structure and interoperability for information retrieval. Published in two parts (Part 1: Thesauri for Information Retrieval, 2011; Part 2: Interoperability with other vocabularies, 2013), it is the standard that governs how controlled vocabularies organize concepts for search, indexing, and cross-vocabulary mapping.

The standard defines:

  • Hierarchical relationships — broader/narrower term (BT/NT) with three subtypes
  • Associative relationships — related term (RT) and its specialized forms
  • Equivalence relationships — preferred term (USE) and used-for (UF)
  • Mapping relationships — cross-vocabulary alignment (exact, close, broad, narrow, related match)
  • Concept scheme structure — how concepts are grouped into thesauri

Key concept: the three hierarchical subtypes

ISO 25964 refines the simple broader/narrower relationship into three subtypes — a crucial distinction that most thesaurus implementations miss:

  1. Generic (BTG/NTG) — genus–species: “mammal” BTG “dog”
  2. Partitive (BTP/NTP) — whole–part: “car” BTP “wheel”
  3. Instantial (BTI/NTI) — class–instance: “planet” BTI “Earth”

This refinement is essential for precise terminology work — the relationship between “car” and “wheel” is fundamentally different from the relationship between “vehicle” and “car”.


How Glossarist implements it

SKOS concept schemes

ISO 25964 Part 2 aligns with SKOS (the W3C Simple Knowledge Organization System). Glossarist implements 14 SKOS concept schemes, defined as taxonomy TTL files in the concept-model repository:

  • Concept groups (dataset-level organization)
  • Concept schemes (vocabulary-level grouping)
  • Mapping relationships (cross-vocabulary alignment)

Hierarchical relationships

The concept hierarchy in Glossarist implements the full ISO 25964 three-subtype model:

UML class diagram of the Glossarist Concept model showing the relationship hierarchy
Concept model hierarchy. ManagedConcept and LocalizedConcept types, their relationship collections, and how typed semantic links (broader/narrower with generic, partitive, and instantial subtypes) connect concepts into navigable systems.

Glossarist implements all three ISO 25964 hierarchical subtypes:

ISO 25964Glossarist typeDirection
BT (broader term)broaderconcept → parent
NT (narrower term)narrowerconcept → child
BTG (broader generic)broader_genericgenus direction
NTG (narrower generic)narrower_genericspecies direction
BTP (broader partitive)broader_partitivewhole direction
NTP (narrower partitive)narrower_partitivepart direction
BTI (broader instantial)broader_instantialclass direction
NTI (narrower instantial)narrower_instantialinstance direction

Mapping relationships

ISO 25964 / SKOS defines five mapping types for aligning concepts across different vocabularies:

SKOS propertyGlossarist typeMeaning
skos:exactMatchexact_matchConcepts are exactly equivalent
skos:closeMatchclose_matchConcepts are approximately equivalent
skos:broadMatchbroad_matchTarget concept is broader
skos:narrowMatchnarrow_matchTarget concept is narrower
skos:relatedMatchrelated_matchTarget concept is related (not hierarchical)

These mappings enable interoperability between Glossarist-managed registries and external vocabularies (e.g., mapping an ISO/TC 211 concept to a corresponding INSPIRE concept).

Full relationship type coverage

Glossarist implements 52 typed semantic relationship types spanning five standards. The ISO 25964 contribution is the hierarchical and associative system:

CategoryStandardTypes
Hierarchical — genericISO 25964 / SKOSbroader, narrower, broader_generic, narrower_generic
Hierarchical — partitiveISO 25964 / ISO 19135broader_partitive, narrower_partitive, has_part, is_part_of
Hierarchical — instantialISO 25964 / ISO 19135broader_instantial, narrower_instantial, instance_of, has_instance
EquivalenceISO 10241-1 / SKOSequivalent, exact_match
SKOS mappingSKOSclose_match, broad_match, narrow_match, related_match
AssociativeISO 10241-1 / ISO 25964see, references, related_concept, related_concept_broader, related_concept_narrower
SpatiotemporalISO 25964 / TBXsequentially_related, spatially_related, temporally_related
LifecycleISO 10241-1 / ISO 19135deprecates, deprecated_by, supersedes, superseded_by, replaces, replaced_by, invalidates, invalidated_by, retires, retired_by
LexicalISO 12620 / TBXhomograph, false_friend
ComparativeISO 10241-1compare, contrast
Designation-levelISO 10241-1abbreviated_form_for, short_form_for

Cross-dataset navigation

Relationships can link concepts across datasets using URN-based references. Each dataset declares a urn (e.g., urn:oiml:pub:v:1:2022 for VIML 2022) in its register.yaml. Cross-dataset relationships reference the target by URN and concept identifier, enabling federated concept systems that span multiple registries.

Inverse edge derivation

Many relationship types are directional — broader in one direction implies narrower in the reverse. The concept-browser derives these inverse edges automatically, so a relationship need only be declared in one direction.

SKOS / RDF export

Glossarist exports concepts as SKOS/RDF via glossarist-ruby:

# Export a GCR package as SKOS Turtle
glossarist convert --input concepts/ --output glossary.ttl --format ttl

The exported RDF file uses standard SKOS properties (skos:Concept, skos:broader, skos:exactMatch, etc.) and can be loaded into any RDF triplestore or SPARQL endpoint.