Evaluating Artificial Intelligence (AI) Models For Library Resource Classification: A Comparative Study Of ChatGPT And DeepSeek

• Sugabsen Martins

  Abdurrahman Ghaji Library, Adamawa State University, Mubi, Nigeria

  Author

• Agwu Patrick Uzodinma

  Kenneth Dike Library, University of Ibadan, Nigeria

  Author

• Kauna Boniface Lambau

  Department of Library and information Science, Faculty of Social Science, Communication and Media Studies, Adamawa State University, Mubi, Nigeria

  Author

This study investigates the accuracy and effectiveness of two prominent Artificial Intelligence (AI) tools; ChatGPT and DeepSeek in applying classification principles to library resources. With the growing integration of AI in bibliographic tasks, it becomes essential to evaluate the reliability of these tools in adhering to standardized classification systems, particularly the Library of Congress Classification (LCC) scheme and author cutter number rules. Using experimental research design, 25 bibliographic titles were subjected to classification by both ChatGPT and DeepSeek, and the outputs were assessed against official Library of Congress Classification (LCC) standards using a scoring rubric. Results revealed that ChatGPT achieved a higher alignment with established classification norms, correctly matching 48% of LCC call numbers and demonstrating superior consistency in cutter number assignment. DeepSeek showed promise but lagged behind, with only 24% exact alignment. Statistical analysis via paired samples test confirmed a significant difference in performance between the two models. While both tools show potential for supporting cataloguing activities, the study concludes that AI classification tools should be used cautiously and under professional supervision. Recommendations were made to enhance AI training using domain-specific data and to promote hybrid AI-human cataloguing approaches for improved accuracy. 

• Download PDF