Authority File

The defining feature of an authority file is the establishment of one preferred term for each entity. This eliminates ambiguity by designating a single authorized access point, while cross-references guide users from variant forms. For example, the Library of Congress Name Authority File (LCNAF) designates 'Twain, Mark, 1835-1910' as the canonical heading, with a see reference from 'Clemens, Samuel Langhorne, 1835-1910'. This principle ensures that all works by an author are collocated under one heading regardless of how the name appears on individual publications.

Authority files employ a rigorous syndetic structure of cross-references to connect related concepts:

• See references (4XX fields): Direct users from non-preferred terms to the authorized heading. Example: 'Kafka, Franz' see 'Kafka, Franz, 1883-1924'.
• See-also references (5XX fields): Link to related authorized headings. Example: 'Machine learning' see also 'Deep learning', 'Neural networks', 'Supervised learning'. This network of associative and equivalence relationships transforms a flat list into a navigable semantic web.

Each record in a modern authority file is assigned a persistent, unique identifier that remains stable even if the preferred heading changes. The Library of Congress Control Number (LCCN) and the Virtual International Authority File (VIAF) ID are prime examples. These identifiers enable entity resolution across disparate systems, allowing a knowledge graph to assert with confidence that two different strings refer to the same real-world entity. This is the foundation of linked data and sameAs linking in semantic web ontologies.

Every heading in a rigorous authority file is justified by literary warrant—the principle that a term's form is derived from the actual usage found in the materials being cataloged, not from a theoretical ideal. Authority records include source citations (670 fields) documenting where the preferred form was found. For instance, a personal name heading cites the author's published works as the source of the name form. This evidentiary basis ensures the file is grounded in objective, verifiable data rather than subjective editorial judgment.

To resolve homography and ambiguity, authority files include scope notes that define the precise meaning and boundaries of a heading. A scope note for 'Mercury' would disambiguate between the planet, the Roman god, the chemical element, and the automobile brand. These notes often include biographical or contextual data—birth and death dates for persons, geographic coordinates for places, founding dates for corporations—that serve as disambiguating attributes for entity resolution algorithms and human catalogers alike.

Authority files encode broader terms (BT) and narrower terms (NT) to establish taxonomic hierarchies, alongside related terms (RT) for non-hierarchical associations. In a subject authority file like the Getty Art & Architecture Thesaurus, 'Baroque architecture' has a broader term 'Architecture by style' and a related term 'Baroque art'. This structured thesaurus logic enables faceted search, query expansion, and the automatic inference of parent-child relationships in knowledge graphs and retrieval-augmented generation systems.