首页|期刊导航|中国中医药图书情报杂志|融合大语言模型与知识图谱的绍派伤寒智能问答系统构建研究

融合大语言模型与知识图谱的绍派伤寒智能问答系统构建研究OA

Study on Intelligent Question-answering System for Shao School of Cold-induced Disease Based on Large Language Models and Knowledge Graph

中文摘要英文摘要

目的 构建一个融合大语言模型与知识图谱技术的中医绍派伤寒学术流派智能问答系统.方法 以绍派伤寒医案为数据来源,利用大语言模型技术自动化抽取医案中包含的实体与关系,构建以"病-证-症-因-治-药"为核心的多维度关联映射的绍派伤寒知识图谱.通过语义映射的两阶段查询转换机制,将用自然语言形式的问题转换为精确的Cypher查询语句,并将离散的查询结果转换为语义连贯、符合专业规范的自然语言回答.结果 本文构建的绍派伤寒知识图谱包含10种实体类型、3 711个节点,10种关系类型、14 940条边,采用Neo4j图数据库实现知识的规范化存储.基于该知识图谱开发的绍派伤寒智能问答系统经专家评估,综合得分为85.02分,高于仅使用大语言模型(82.04分)和仅知识图谱查询结果(59.46分).结论 绍派伤寒智能问答系统可实现高效知识检索与兼具专业性与可解释性的回答生成,可促进绍派伤寒学术体系的数智化传承与创新,其构建思路可为其他智能诊疗系统提供参考和借鉴.

Objective To construct an intelligent question-answering system for Shao School of cold-induced disease by integrating large language models(LLMs)and knowledge graph(KG)technology.Methods Utilizing medical cases of Shao School of cold-induced disease as the data source,an LLMs-based approach was implemented to automatically extract entities and relationships contained in medical cases of Shao School of cold-induced disease,thereby constructing a multi-dimensional knowledge graph hierarchically mapped around the core axes of"disease-pattern-symptom-cause-treatment-herb".A semantic mapping-based,two-stage query transformation mechanism was proposed to convert natural language queries into precise Cypher graph queries.Concurrently,the retrieved graph data were transformed into coherent,professional standardized natural language answers.Results The newly constructed knowledge graph of Shao School of cold-induced disease contained 10 entity types,3 711 nodes,10 relationship types and 14 940 edges,with standardized storage achieved via Neo4j graph database.Expert evaluation demonstrated that the developed intelligent question-answering system achieved a comprehensive score of 85.02,significantly outperforming the standalone LLMs(82.04)and the direct KG query results(59.46).Conclusion The developed intelligent question-answering system of Shao School of cold-induced disease enables efficient knowledge retrieval and generates highly professional and interpretable answers.This framework effectively promotes the digital and intelligent preservation and innovation of the academic system of Shao School of cold-induced disease,offering references for the construction ideas of other intelligent clinical diagnosis and treatment systems.

金海心;吴毅晗;何素珍;张越;赵佳璐

浙江中医药大学,浙江 杭州 310053浙江中医药大学,浙江 杭州 310053浙江中医药大学,浙江 杭州 310053浙江中医药大学,浙江 杭州 310053浙江中医药大学,浙江 杭州 310053

医药卫生

绍派伤寒大语言模型知识图谱智能问答系统

Shao School of cold-induced diseaselarge language modelknowledge graphintelligent question-answering system

《中国中医药图书情报杂志》 2026 (5)

32-38,7

国家级大学生创新创业训练计划(202410344004)浙江中医药大学校级科研基金(2023JKJNTS23)

10.3969/j.issn.2095-5707.202508085

评论