Glossary Updates16 new terms added to the glossaries · November 12, 2026, 21:23 CET
AI TechDocKnowledge
  • English (US)
  • English (UK)
  • Deutsch
All context cards

Context cardTransitional topic

What concept theory offers ontology work

What can ontology and knowledge graph work take from research on how concepts are represented?

By knowledge.aitechdoc.world

Reviewed

Review log and changes

The short answer

Three distinctions that common models leave implicit: what is definitional and what is only typical, how values can be compared, and how one occurrence is described on several levels. Frames, conceptual spaces, prototypes and cascades give these distinctions names. They are lenses for modeling and review — not a modeling language and not a substitute for an agreed vocabulary and its governance.

For: Ontology modelers, knowledge engineers, information architects and technical writers

Key points

  • Definitional vs. typical: prototype research and the decomposition debate show that not every expected property belongs to a definition.
  • Comparable values: conceptual spaces model value ranges and similarity, which schemas usually leave to the application.
  • One occurrence, several levels: cascades relate descriptions of the same event by what they amount to in context — a relation standard vocabularies rarely provide.
  • Frames make attributes and constraints explicit and resemble classes with typed properties, but they model cognition, not data exchange.
  • Use the lenses to review and evolve a model; keep the model itself in its own language and under its own governance.

The context

Why look at concept theory at all

An ontology or a knowledge graph is a set of decisions about concepts: which classes exist, which properties they have, what must hold and what may vary. Research on concept representation — collected, for example, in the open-access volume Concepts, Frames and Cascades in Semantics, Cognition and Ontology (2021) — studies how people make the same decisions. Its findings do not tell a modeler what to build, but they explain where models and readers part ways.

Three lenses

1. Definitional or typical?

Prototype research shows that categories have typical members and default expectations; the debate between decomposition and atomism shows how hard it is to say which features belong to a meaning. In a model, this suggests marking which properties are necessary and which are defaults that instances may override — and expecting the defaults to shift first (semantic drift).

2. How do values compare?

Conceptual spaces describe values as points in dimensions with a distance. Most information models declare a datatype and perhaps a range, but not when two values count as similar. Where classification, search or alignment depends on similarity, the measure is part of the model’s meaning and worth stating.

3. One occurrence, several levels

Cascades (Sebastian Löbner, after Alvin Goldman) describe how one act is correctly categorized on several levels at once: pressing a button, starting a machine, beginning a production run. Taxonomies (subclass) and part–whole relations do not capture this; it is a relation between descriptions of one event, and different roles care about different levels. Models that serve several audiences profit from naming the level a statement is made on.

Limits

Frames and record types (Robin Cooper’s Type Theory with Records) come close to typed schemas, but concept theory remains a theory of cognition and meaning. Ontology languages such as RDF-based vocabularies have their own semantics, and shared vocabularies need agreement and governance that no theory of concepts supplies. The lenses help to review and evolve a model; they never imply that a model is correct, complete or compliant.

The subject area Concepts, frames and cascades in the encyclopedia covers each theory with its sources.

Questions readers ask next

Is this an ontology method?
No. It is a set of questions taken from research on concepts, to be used when designing or reviewing a model. The model is still built and governed with the usual methods and languages.
Why should defaults be marked?
Because people treat typical properties as if they were part of the concept. A model that separates necessary properties from defaults is easier to align with other vocabularies and to revise when usage changes.
Where do cascades show up in documentation?
Wherever one act matters on several levels for different readers — the operator’s action, the machine function it triggers, the production step it begins. Naming the level avoids mixing statements that are true at different levels.

Sources

  1. Sebastian Löbner et al. (eds.), Concepts, Frames and Cascades in Semantics, Cognition and Ontology (open access, CC BY 4.0) — Springer, January 1, 2021
  2. Sebastian Löbner, Thomas Gamerschlag, Tobias Kalenscher, Markus Schrenk and Henk Zeevat, Introduction (in: Concepts, Frames and Cascades in Semantics, Cognition and Ontology, open access, CC BY 4.0) — Springer, January 1, 2021
  3. Sebastian Löbner, Cascades. Goldman’s Level-Generation, Multilevel Categorization of Action, and Multilevel Verb Semantics (in: Concepts, Frames and Cascades in Semantics, Cognition and Ontology, open access, CC BY 4.0) — Springer, January 1, 2021
  4. Robin Cooper, Perception, Types and Frames (in: Concepts, Frames and Cascades in Semantics, Cognition and Ontology, open access, CC BY 4.0) — Springer, January 1, 2021
  5. Corina Strößner, Annika Schuster and Gerhard Schurz, Modification and Default Inheritance (in: Concepts, Frames and Cascades in Semantics, Cognition and Ontology, open access, CC BY 4.0) — Springer, January 1, 2021

Review log and changes

Every context card is checked against its sources before it is published, and again whenever it changes; the date under the byline is the last review. Corrections (something was wrong) and additions (something was missing) are logged below with date and time (Berlin time). Typos, formatting and link fixes are not listed.

Reviewed

No corrections or additions since publication.