Using Generative AI for Bibliographic Description: A Study with ChatGPT 4

This study explores the use of Generative Artificial Intelligence (GAI) in bibliographic description in university library catalogs. It focuses on GAI's potential in enhancing efficiency and maintaining consistency of bibliographic description while complying with cataloging standards such as MARC21, AACR2 and RDA. The study followed a qualitative methodology, examining 10 use cases in metadata extraction, RDA compliance of GAI generated catalog records, and error checking of existing catalog records. The study used OpenAI’s ChatGPT 4 to execute the practical tests. The study concludes that GAI can significantly assist in improving cataloging efficiency. It was highly satisfactory in error correction of existing MARC21 records. GAI was also a prospective tool for educating novice librarians on the use of MARC21 and RDA standards. CATMELK, a custom GPT model created with ChatGPT 4 could effectively convert images from books’ title, copyright and cover pages into RDA compliant MARC21 records. However, when the tests were conducted in January 2024, ChatGPT 4 faced difficulties in direct conversion of Sinhala and Tamil data from images to MARC2 1 records. Considering the rapid development of GAI technology and applications, it is recommended to repeat the tests on a continuing basis.