基于中文医疗知识图谱的智能问答系统设计与实现方法

Design and Implementation of Intelligent Q&A System Based on Chinese Medical Knowledge Graph

目的:构建基于中文医疗知识图谱的智能问答系统,使人们通过人机交互的方式就能完成简单的自我诊疗。方法:通过词性标注的方法获取用户提出问句中的医疗实体,再利用结合基于共享层的卷积神经网络(SH-CNN)与词频-逆文本频率(TF-IDF)算法的混合算法来计算出系统中与问句语义最接近的问题模板。最后根据获取问题模板的问句类型以及问句中的医疗实体构建cypher语句,从知识图谱中检索答案返回给用户。结果:该系统具有较强的问题解答能力,回答准确率达90.7%。结论:基于医疗知识图谱的问答系统为用户提供了快速准确的答案,可在一定程度上缓解医疗资源紧缺的矛盾,是医疗领域信息化的必然趋势。

Objective:To construct an intelligent Q&A system based on Chinese medical knowledge graph,help people to complete simple self-diagnosis and treatment through human-computer interaction.Methods:Using the method of part of speech tagging to obtain the medical entities in the questions raised by users,and then to calculate the problem template which is closest to the question semantics in the system by combining the hybrid algorithm(SH-CNN)and(TF-IDF)based on Shared layer.Finally to construct the cypher statements according to the question types of the question templates and the medical entities in the questions,and to retrieve the answers from the knowledge graph.Results:The system has a strong ability to answer questions,and the answer accuracy is 90.7%.Conclusion:The Q&A system based on the medical knowledge graph provides users with fast and accurate answers,which can alleviate the contradiction of the shortage of medical resources to a certain extent,and is an inevitable trend of the medical informationalization.