ARTIFICIAL INTELLIGENCE IN CATALOGING ARTISTS' BOOKS: experiments with CatalogerGPT
Year: 2026
Type: conference-paper
Abstract: This research analyzes the use of artificial intelligence in the cataloging of artists' books through experiments with CatalogerGPT, comparing the results obtained with human cataloging carried out by the Brazilian National Library Foundation. The study uses the work Vaslav Nijinski: SOU, by Amador de Carvalho Perez, as its object, considering the specificities of artists' books as both bibliographic and artistic works. This is an applied research study with a qualitative approach and a comparative method. A specific prompt based on Anglo-American Cataloguing Rules, Second Revised Edition (AACR2r) and artists' book cataloging practices was developed, requesting MARC 21 records, subject indexing, and detailed description. The results demonstrated that CatalogerGPT performed well in descriptive and subject representation, including more detailed notes and a greater number of access points than human cataloging. However, errors were also identified in specific fields, as well as limitations related to the need for digitization of the work and human review for data validation. Conference abstract published in the proceedings of TOI2026 — XII Congresso Internacional em Tecnologia e Organização da Informação. ISSN 2965-4130. Original publication in Doity
Language: Portuguese
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7216282938