Demonstrating the Development of an AI-powered “Ask a Librarian” Service via LLM Fine-tuning on Low-end Hardware for the Library
DESIDOC Journal of Library & Information Technology · 13.11.2026 · 46 (6) · 722-734
DOI: 10.14429/djlit.22101 ↗Kurz gefasst
Fachartikel in der Zeitschrift DESIDOC Journal of Library & Information Technology, online erschienen am 13.11.2026. Verfasst von Kaustav Kumar Khanikar und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 2.5).
Abstract (Originalsprache)
This paper presents the process of developing an AI-powered Virtual Library Assistant (ViLA) prototype that can automate the “Ask a Librarian” service. The paper showcases a scenario for the practical implementation and integration of AI within a library setting. The methodology entails utilising low-end hardware and an open-source pretrained model, which is then fine-tuned on a curated dataset of FAQs and common questions typically answered by the library. By employing Low-Rank Adaptation (LoRA), the researchers were able to fine-tune the model efficiently by updating only a small subset of weights rather than the entire architecture. After fine-tuning, the model is served using the Streamlit library of Python, which provides an interface for users to interact. The fine-tuned model achieved strong performance with a BERTScore Precision and recall of 0.8499 and 0.8413 respectively. The system further received a mean rating of 3.86 (out of 5) and a median rating of 4.0 on human evaluation. The system is not a replacement for a librarian; rather, it acts as a hybrid, human-in-the-loop service model, improving scalability while maintaining the professional integrity of support. This project demonstrates how librarians can utilise AI to deliver more effective and responsive user services.
Zitieren
APA
Khanikar, K. K., & Barua, N. (2026). Demonstrating the Development of an AI-powered “Ask a Librarian” Service via LLM Fine-tuning on Low-end Hardware for the Library. DESIDOC Journal of Library & Information Technology, 46(6), 722-734. https://doi.org/10.14429/djlit.22101
Vancouver
Khanikar KK, Barua N. Demonstrating the Development of an AI-powered “Ask a Librarian” Service via LLM Fine-tuning on Low-end Hardware for the Library. DESIDOC Journal of Library & Information Technology. 2026;46(6):722-734. doi:10.14429/djlit.22101
BibTeX
@article{Khanikar2026,
author = {Khanikar, Kaustav Kumar and Barua, Niraj},
title = {Demonstrating the Development of an AI-powered “Ask a Librarian” Service via LLM Fine-tuning on Low-end Hardware for the Library},
journal = {DESIDOC Journal of Library & Information Technology},
year = {2026},
volume = {46},
number = {6},
pages = {722--734},
doi = {10.14429/djlit.22101}
}
EUVIMED