Controlled Vocabularies: Synonym Rings and Authority Files

Natural language creates two problems for information spaces: ambiguity, where the same word means different things, and synonymy, where different words mean the same thing. A controlled vocabulary is a well-defined subset of language that solves this by establishing preferred terms for concepts, along with their variant, broader, narrower, and related terms. Choosing preferred terms draws on three warrants: literary warrant (how the term appears in existing literature), user warrant (how your actual audience talks), and organizational warrant (how your own organization refers to it).

Synonym rings are the simplest form: just a list of equivalent terms for a concept, used to broaden what a search retrieves, importantly: they're not used for indexing content, only for expanding queries at search time. Authority files are the next step up: an index of authority records, each defining one preferred term along with its cross-referenced variants and a note validating why that term is authoritative. They come in two flavors, name authority files, for unique entities like people or organizations, and subject-heading authority files, for standard topic terms.

A synonym ring just widens the net at search time. An authority file goes further, it actually declares which term is correct, and keeps a record of why.

Exercise

The scenario: A large retail chain is experiencing two distinct search problems. First, customers searching for products using brand names, generic names, common misspellings, and regional variants get inconsistent results, a search for "trainers" and a search for "sneakers" don't reliably return the same products. Second, the product catalog uses inconsistent names for the same supplier organizations across internal records, causing reporting failures when someone tries to total purchases from a single supplier.

Which tool is the better fit for the first problem, inconsistent product search results across synonymous terms?
Which tool is the better fit for the second problem, inconsistent supplier names breaking internal reporting?