现代情报2026，Vol.46Issue(1)：173-186,14.DOI:10.3969/j.issn.1008-0821.2026.01.015
融合DeepSeek-R1和RAG技术的先秦文化元典智能问答研究
Research on Intelligent Question Answering for Pre-Qin Cultural Classics by Integrating DeepSeek-R1 and RAG Technologies
摘要
Abstract
[Purpose/Significance]As the source literature of Chinese civilization,the pre-Qin cultural classics contri-bute to providing historical evidence and value judgments for building a modern Chinese national civilization and enhancing national cultural soft power through knowledge organization and intelligent application.This study aims to develop an inte-lligent Q&A system for pre-Qin cultural classics based on Retrieval-Augmented Generation(RAG)technology to promote the intelligent application and inheritance of relevant knowledge.[Methods/Process]Taking the"Three Commentaries on the Spring and Autumn Annals"published by Zhonghua Book Company as the research object,the research constructed an ontology model for pre-Qin cultural classics,and used DeepSeek-R1 for knowledge extraction,and constructed a knowledge graph.Based on the LangChain framework,four Retrieval-Augmented Generation(RAG)methods-GraphRAG,NaiveRAG,LightRAG,and HybridRAG-were employed to enhance the retrieval ability of the large language model,and the question-answering ability was evaluated from both quantitative and mixed aspects.[Result/Conclusion]The research results show that DeepSeek-R1 demonstrates excellent extraction performance,with the generated triples effectively covering key knowledge while maintaining high quality.In the intelligent question-answering evaluation,different RAG approaches have their respective strengths and weaknesses.GraphRAG performs well across various question types and evaluation dimensions,particularly excelling in verification-and-traceability-oriented and applied-practice-oriented questions.NaiveRAG shows better performance in factual knowledge-oriented questions.Based on comprehensive quantitative and hybrid evaluations,selecting appropriate RAG technology according to practical application scenarios is crucial.关键词
先秦文化元典/大语言模型/DeepSeek/检索增强生成/智能问答Key words
pre-Qin cultural classics/large language models/DeepSeek/Retrieval-Augmented Generation/intelli-gent question answering分类
信息技术与安全科学引用本文复制引用
张强,高颖,任豆豆,韩牧哲,包平..融合DeepSeek-R1和RAG技术的先秦文化元典智能问答研究[J].现代情报,2026,46(1):173-186,14.基金项目
国家社会科学基金青年项目"出土文献的多模态知识组织与融合研究"(项目编号:23CTQ038). （项目编号:23CTQ038）
