Comparative evaluation reveals insufficient accuracy and consistency of ChatGPT in generating Dewey Decimal Classification numbers, highlighting the necessity of cataloguer oversight.
This study investigates the potential use of large language models (LLMs) for generating Dewey Decimal Classification (DDC) call numbers. The authors used ChatGPT to generate DDC call numbers for monographs based on their bibliographic metadata and compared these with numbers built ‘manually’ by cataloguers using WebDewey. The collection of materials reclassified was previously located in DDC class 572 – which used to refer to anthropology, ethnology, or race but now refers to biochemistry – and were identified for reclassification as part of reparative project work. Although ChatGPT did not achieve sufficient accuracy, consistency, or transparency to justify continued use for DDC reclassification at this time, the study highlights broader issues relating to AI reliability, bias, transparency, and the continuing importance of human expertise in library knowledge organisation as well as a need for human oversight in culturally sensitive library work. The overall experience was positive as the authors identified several professional and organisational benefits from undertaking the process.
No takes yet. Share an insight, caveat, or question.
Steele et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: