Application Research on Intelligent Reference Services in Libraries Based on Large Language Models

Abstract

As the digital information environment grows increasingly complex and users' knowledge needs deepen, traditional library reference services face efficiency bottlenecks when addressing open-domain, multi-turn, and semantically complex inquiries. Large language models, built on the Transformer architecture, demonstrate profound capabilities in deep semantic understanding and generation, offering a new technological pathway for reshaping the paradigm of reference services. This study aims to systematically explore the theoretical foundation for integrating large language models with library reference services, construct a technically layered and decoupled architecture for an intelligent reference system equipped with domain adaptation and knowledge enhancement capabilities, and elucidate its core operational mechanisms. The research provides a detailed analysis of domain knowledge integration methods based on Retrieval-Augmented Generation and fine-tuning of large models, as well as context understanding mechanisms that support multi-turn dialogues and the integration of multi-source information. Furthermore, the paper discusses future directions for the evolution of intelligent reference systems from three dimensions: personalized services, the reconstruction of human-computer collaboration processes, and trustworthy assurance systems. This study offers a systematic theoretical framework and technical design concepts for libraries to leverage large language models in building a new generation of intelligent knowledge service infrastructure.