Controlled vocabularies are staff-maintained, unlike folksonomies

What it is

A controlled vocabulary has named people ruling on preferred terms and relations, with a process and an owner. A folksonomy is tags users spontaneously stick on objects: anyone can coin a word, and there is no single preferred term. Both can make content findable; they are not the same kind of system. The first wants stability and aggregability; the second wants low cost and a fit to current wording. Promoting user tags straight into navigation classes treats a folksonomy as a controlled vocabulary. Forbidding people to search with their own words treats the controlled vocabulary as the only legal language.

Different natures mean different governance, UI, and failure modes. One “tag management” screen cannot paper over the split.

Why it happens

The bottleneck on the controlled side is ruling speed: the domain moves, the vocabulary lags, and content is stuck to a stale preferred term. The bottleneck on the folk side is uncontrolled synonymy and polysemy: ten users, ten words on one object, aggregation fails; one word used for two things, filters lie. Control buys machine-computable consistency from a few people. Folksonomy buys coverage from many people in the moment, and patches aggregation with later clustering.

They can join: folk words enter the controlled process as non-preferred candidates and become preferred terms or official aliases only after admission. Without that gate, the folk lexicon floods the nav and control collapses at once.

Studying it

Measure consistency and freshness of the two systems separately; do not mash them into one “tag quality” score.

• Paradigms: the same objects, expert controlled indexing versus user tags; compute tag agreement, time lag, retrieval success. Then audit whether folk words were wrongly written into navigation.
• Independent variables: who may coin terms, whether a coinage appears immediately in filters, presence of a review gate.
• Dependent variables: split of tags on one object, rate of one word meaning two things, delay from a new wording to official retrievability, times the nav is polluted by folk terms.
• Methodological note: enthusiast communities show higher folksonomy agreement than ordinary consumers. Match the sample to the product’s users. An expert gold standard will bias toward controlled wording; when scoring folk tags, the criterion is whether a user query hits, not whether the tag equals the preferred term.

Where it stops holding

Specialist collections (medical subject headings, legal topics) almost must be controlled; folk tags are at most personal organization and cannot drive public nav. Pure social bookmarking can live on folksonomy alone; a controlled vocabulary is over-governance there. Inside a company, “anyone can add a category in the CMS” looks controlled and is an ownerless folksonomy. Treat that failure as folk: shut ungoverned minting first, then talk vocabulary—do not train people to “fill it in carefully.”

Applying it

• Declare two pipes: nav, filters, and official retrieval consume only preferred terms; user tags stay on the object, may drive personal filters, and do not enter global facets by default.
• High-frequency folk words regularly enter a controlled candidate list; an owner decides promote, alias, or remain folk.
• Forbid unreviewed “add a category” in the CMS. That is not agility; it turns the vocabulary into a folksonomy.
• Verify with a newly fashionable wording. Users should be able to find their own objects with it immediately; global nav and official filters should not show it before a ruling. Reverse those two, and the pipes are wired backwards.

Related

• Within the group: G1.10.1 A controlled vocabulary gives each concept one preferred term; others point to it · G1.10.2 A thesaurus describes hierarchy and association with broader, narrower, and related terms · G1.10.3 Authority files hold canonical names for entities and resolve homographs and aliases · G1.10.5 At retrieval, non-preferred terms map to the preferred term and yield the same results · G1.10.6 Vocabularies need versioning and change records as the domain evolves
• Adjacent: G1.04 Labeling systems · G1.11 Architectural scalability and evolution · G1.07 Content inventory and audit
• Search terms: folksonomy · controlled vocabulary · user-generated tags