图书馆信息检索:从MARC到RAG  

Library Information Retrieval:From MARC to RAG

作　　者：王诗卉 蔡丹丹 Wang Shihui;Cai Dandan(Shanghai Library(Institute of Scientific and Technical Information of Shanghai))

机构地区：[1]上海图书馆(上海科学技术情报研究所)

出　　处：《信息与管理研究》2024年第6期34-43,共10页Journal of Information and Management

摘　　要：图书馆信息检索经历了从手动卡片检索、电子化目录检索到现代全文检索的演变。同时,图书馆馆藏资源的形态也发生了巨大变化。从最初的物理馆藏,到逐步引入各类数据库,以及到如今涵盖各类模态资源的数字资源体系。进入大模型时代,图书馆信息检索开始迈向智能化。大模型能够处理复杂数据来源的自然语言查询并扩展检索的方式。本文在回顾历史检索技术发展的基础上,探讨了大模型对于图书馆信息检索的提升,提出了一个基于RAG的图书馆信息检索系统架构,分析了未来大模型对图书馆信息检索模式的影响。The evolution of library information retrieval has progressed from manual card retrieval and electronic catalog retrieval to full-text retrieval.Simultaneously,the forms of library collections have undergone significant transformations,transitioning from physical collections to various databases and,ultimately,to a digital resource system encompassing diverse modalities.In the era of large language models,library information retrieval is advancing towards intelligence.Large language models can handle natural language queries from complex data sources,enhancing the retrieval methods.Based on a review of the historical development of retrieval technologies,this paper explores the enhancements brought by large language models to library information retrieval,proposes a library information retrieval system architecture based on Retrieval-Augmented Generation(RAG).It also analyzes the impact of large language models on future library information retrieval patterns.

关 键 词：图书馆信息检索 大模型 RAG 人工智能 

分 类 号：G252.7[文化科学—图书馆学] G250.7