摘要
为提升以大语言模型为代表的人工智能技术在图书馆参考咨询与知识服务中的应用性能与领域适配性,文章提出融合多路检索的检索增强生成智能问答系统。文章通过深度融合OPAC的书目数据与来自公共查询门户的摘要信息等多源数据,构建一个大规模、高质量的动态知识库;集成多路检索与提示学习技术,设计高效、专业问答框架,以提升领域知识检索正确率及答案生成质量。实验结果表明,该系统在问答正确率上较传统方法与通用大模型提升21%,ROUGE-1提升约35%,有效验证了其在图书馆知识服务领域的卓越性能与适配优势。
To improve the performance and domain adaptability of artificial intelligence(AI)technologies represented by large language models(LLMs)in library reference and knowledge services,this article proposes an intelligent question-answering(QA)system that incorporates multi-path retrieval with Retrieval-Augmented Generation(RAG)technology.The study presents a large-scale,high-quality dynamic knowledge base by deeply integrating OPAC bibliographic data with abstract information from public query portals and other multisource data.Furthermore,it develops an efficient,professional QA framework that integrates multi-path retrieval with prompt learning techniques to enhance the accuracy of domain knowledge retrieval and the quality of answer generation.Experimental results demonstrate that the proposed system enhances QA accuracy by 21%compared to traditional methods and general-purpose large models,and increases ROUGE-1 scores by 35%,effectively validating its superior performance and adaptability in the field of library knowledge services.
出处
《图书馆论坛》
北大核心
2025年第12期130-139,共10页
Library Tribune
关键词
图书馆智慧服务
大语言模型
检索增强生成
问答系统
intelligent library service
large language model
retrieval-augmented generation
QA system
