Name Authority Control Evaluation

Collocation is the gathering of all of a single identity's works under one controlled access point; it fails when a person or body is split across several variant forms, so a searcher finds only part of their output. Disambiguation is the separation of distinct identities that share a name; it fails when different people are conflated under one heading, so a search retrieves a mixture. They are the recall and precision sides of authority control respectively, and a good evaluation measures both, because an authority file can be strong on one and weak on the other.

FRAD (Functional Requirements for Authority Data) provides the conceptual yardstick. It defines the entities authority data describes — persons, families, corporate bodies, names, identifiers — and the user tasks authority control must support: find, identify, contextualize, and justify. Evaluation uses these as criteria: it asks whether the authority data lets users find and unambiguously identify an entity, place it among related entities, and see the justification for the chosen access point. This turns 'is the authority file good?' into measurable questions tied to the tasks the file exists to serve.

External identifier systems provide independent, persistent anchors for real-world identities, which helps both in establishing ground truth and in assessing interoperability. Comparing a file's headings against VIAF clusters or matching creators to ORCID and ISNI records can reveal splits and conflations and shows how well the catalogue connects to the wider identity ecosystem. But the linkage must be verified rather than trusted blindly: an evaluation should confirm that the identifier actually denotes the intended entity, since incorrect identifier matches can introduce the very conflation errors authority control is meant to prevent.