Hannah Moutran
Library Specialist, Implementations of Artificial Intelligence
University of Texas at Austin
At the University of Texas at Austin Libraries, the AI-Assisted Music Cataloging Project demonstrates how libraries can leverage large language models alongside other tools to automate copy-cataloging workflows while maintaining quality control. The system generates metadata from images of CDs and LPs, queries OCLC WorldCat for matching records, and uses a large language model to evaluate matches and assign confidence scores. Confidence scores are then adjusted with automated comparisons of track titles and publication years. The workflow creates a number of cataloging tools, including a batch upload file; a physical sorting spreadsheet; and an optional HTML review interface that allows catalogers to visually compare OCLC records with object images, approve or reject matches, and export their decisions for integration into cataloging files. Institutions can choose their level of verification before proceeding to batch upload, supported by a script that creates bibliographic, holdings, and item records in Alma from OCLC data. The open-source toolkit emphasizes batch processing optimization for cost efficiency and models a replicable method for institutions seeking to responsibly integrate AI into their technical services operations.
