Concept System Types

ISO 704:2022 §5.6.3 classifies concept systems along three orthogonal dimensions. A real-world concept system carries one value from each dimension — e.g. the OIML measurement standard decomposition is generic + multidimensional + monohierarchical. Glossarist models each dimension independently so the type emerges from the data, not from a declarative tag.

The three dimensions

By relation type

Type	What it is	Glossarist shape
Generic	All concepts relate as genus ↔ species (§5.5.4.2.1)	`GenericHyperedge` rakes (see Generic Relations)
Partitive	All concepts relate as whole ↔ part (§5.5.4.3)	`PartitiveHyperedge` rakes (see Partitive Relations)
Associative	All concepts relate thematically (no hierarchy) (§5.5.5)	Binary `related` edges — `see`, `compare`, `related_concept`, etc. (see Relationships)
Mixed	Combination of two or more relation types	All of the above, coexisting on the same `ManagedConcept`

Most real-world terminological vocabularies are mixed. Glossarist’s separation of n-ary hyperedges (per-file at relations/<id>/) from binary edges (inline on the concept) makes the mix explicit in the file layout — see Hyperedges — Per-file storage.

By criterion count

Type	What it is	Trigger
Monodimensional	Superordinate concepts subdivided under one criterion (§5.5.4.2.1 Example 1)	One `*Hyperedge` per comprehensive, OR multiple binary edges without grouping
Multidimensional	Superordinate concepts subdivided under two or more criteria (§5.5.4.2.1 Example 4; §5.6.3)	Two or more `*Hyperedge` files on the same comprehensive, each carrying a distinct `criterion`

Multidimensionality is what hyperedges were designed for — binary edges can’t express “these species belong to this rake, those species belong to that rake” without the criterion dimension. See Generic Relations — Multidimensionality.

The check-partitive-relation-coherence and check-generic-relation-coherence validators warn when a comprehensive carries 2+ rakes and any lack a criterion — without criterion, multidimensionality can’t be verified.

By hierarchy shape

Type	What it is	Glossarist shape
Monohierarchical	Each concept has at most one immediate superordinate (§5.5.4.2.1 Example 1)	Tree-shaped concept graph
Polyhierarchical	One or more concepts have multiple immediate superordinates (§5.6.3 Example 5)	DAG-shaped concept graph (directed acyclic)

Polyhierarchy emerges naturally from cross-dataset related edges — e.g. an OIML G18 term may see a VIM concept (its origin) and also broader an OIML V 2-200 concept (its genus in the OIML hierarchy). Glossarist’s directional related graph supports polyhierarchy without special handling; the concept-browser renders it as a DAG.

Worked example — OIML V 2-200:2010 measurement standard

Type: generic + multidimensional + monohierarchical

Dimension	Value	Evidence
Relation type	generic	Genus/species rakes per §5.5.4.2.1
Criterion count	multidimensional	6 distinct criteria on the same genus (by realization medium, by calibration role, by governance level, by metrological hierarchy, by reference/working, by travel)
Hierarchy shape	monohierarchical	Each species has exactly one genus (`measurement standard`) — no concept is a species of two different genera

In Glossarist this is 6 GenericHyperedge files under relations/viml-5-1/, each carrying its own criterion:

relations/viml-5-1/
  by-realization-medium.yaml      # criterion: { eng: by realization medium }
  by-calibration-role.yaml         # criterion: { eng: by calibration role }
  by-governance-level.yaml         # criterion: { eng: by governance level }
  by-metrological-hierarchy.yaml   # criterion: { eng: by metrological hierarchy }
  by-reference-or-working.yaml     # criterion: { eng: by reference or working }
  by-travel.yaml                   # criterion: { eng: by travel }

See the Generic Relations — canonical computer mouse example for a 2-criterion worked example with delimiting characteristics.

Worked example — optomechanical mouse (mixed system)

Type: mixed + multidimensional + monohierarchical

ISO 704:2022 §5.5.4.2.2 uses optomechanical mouse as both:

A species in a generic rake under computer mouse (criterion: by movement detection)
A comprehensive in a partitive rake decomposing it into physical parts

Dimension	Value	Evidence
Relation type	mixed	Both generic and partitive rakes on the same comprehensive
Criterion count	multidimensional	At minimum: one criterion per rake type
Hierarchy shape	monohierarchical	Each concept has one parent in each rake; cross-rake parents are different roles

Glossarist’s file layout makes the mixed nature explicit:

concepts/optomechanical-mouse.yaml     # status: valid

relations/example-computer-mouse/
  by-movement-detection.yaml           # type: generic_relation

relations/example-optomechanical-mouse/
  physical-component-decomposition.yaml # type: partitive_relation

Choosing a concept system type

If your domain…	Use
Has strict taxonomy (biology, chemistry nomenclature)	Generic + monodimensional + monohierarchical
Decomposes physical objects (anatomy, mechanical assemblies)	Partitive + monodimensional + monohierarchical
Has multiple independent classification axes (metrology, standards)	Generic + multidimensional + monohierarchical
Has overlapping categories (library science, indexing)	Generic + multidimensional + polyhierarchical
Has no hierarchy, just thematic links (collections, catalogs)	Associative (binary edges only)
Combines several of the above (most real domains)	Mixed + multidimensional + polyhierarchical