Creating MyGPT for Cataloging Tasks: The Experience of the Office of Academic Resources, Chulalongkorn University

Abstract

The objectives of this study were to 1) Evaluate the level of accuracy of MyGPT-generated Thai bibliographic records. 2) Find the error patterns that occur repeatedly. 3) Examine the most appropriate method for creating bibliographic records using MyGPT. The method started with the creation of MyGPT with instructions about workflow, well-formed data structures, and the fields required in the records. Then generated ten Thai bibliographic records using MyGPT and compared the results with the records created by a librarian. The study found that MyGPT is still unable to generate fully accurate and instantly usable bibliographic records from images of front matter and other book sections, but entering books’ data directly into MyGPT improved accuracy. The result showed the potential for enhancing MyGPT through the collaboration between librarians and artificial intelligence. For data accuracy and structure, constant fields received a higher score than those requiring complex recording methods. The most common mistakes were compliance with standards such as MARC, RDA, and cutter number assignments.