Lesson 3 — Controlled Vocabularies & Semantic Structure
A controlled vocabulary is a governed set/structure of allowed terms used to classify, tag, browse, sort and retrieve information. Chapter 9 treats it as both Reference Data (managed values) and Metadata (values used to describe other information).
The high-yield family
| Form | Deciding clue |
|---|---|
| Term / pick list | Allowed values; few or no explicit relationships |
| Vocabulary view | A subset of a larger vocabulary for one audience/topic |
| Micro-controlled vocabulary | Enterprise-aligned subset plus governed specialist extensions |
| Synonym ring | Equivalent terms treated equally for retrieval |
| Authority list | One preferred term; variants redirect to it |
| Taxonomy | Controlled classification/hierarchy/facets |
| Thesaurus | Equivalent + hierarchical + associative relationships |
| Ontology | Rich semantic domain model: classes/concepts, instances, attributes, relations, events |
Relationship clues
- Equivalent: same/synonymous meaning.
- Hierarchical: broader/narrower or whole/part.
- Related/associative: meaningful association that is neither equivalent nor hierarchical.
Synonym Ring vs Authority List
If “doctor,” “physician” and “MD” should all work equally → Synonym Ring.
If Physician is the approved label and the others redirect to it → Authority List.
Taxonomy vs Thesaurus vs Ontology
Use Taxonomy when classification/browsing is the need.
Use Thesaurus when broader/narrower, equivalent and related relationships matter.
Use Ontology when the domain itself must be represented with richer machine-interpretable semantics.
Folksonomy
A folksonomy is bottom-up social tagging. It can reveal user language and inform vocabulary design, but it is not automatically an approved enterprise vocabulary.
Alignment principle
Content vocabularies should align where practical with enterprise conceptual definitions. If a content taxonomy defines “Supplier” differently from enterprise data models, structured/unstructured integration becomes harder.
Source: pp. 293–295.