Computer Engineering and Applications ›› 2021, Vol. 57 ›› Issue (22): 304-312.DOI: 10.3778/j.issn.1002-8331.2103-0038

• Engineering and Applications • Previous Articles    

Construction of Question Answering System for Suicide Tendency Detection Based on Knowledge Graph

WU Shuzhao, LI Gongquan, BU Mingwei   

  1. 1.School of Geosciences, Yangtze University, Wuhan 430100, China
    2.School of Geography and Information Engineering, China University of Geosciences, Wuhan 430074, China
  • Online:2021-11-15 Published:2021-11-16

基于知识图谱的自杀倾向检测问答系统构建

武书钊,李功权,卜明伟   

  1. 1.长江大学 地球科学学院,武汉 430100
    2.中国地质大学 地理与信息工程学院,武汉 430074

Abstract:

Due to the people lack of psychological knowledge, the prevalence rate of mental illness is rising rapidly. In order to solve the problem, a small psychological counseling knowledge graph and Question Answering system(QA) is constructed. The system can help consultants to acquire psychological knowledge in time, also can identify the consultants who have suicidal tendency to prevent the occurrence of danger. Firstly, collect some psychological information data to build the dictionary, and generate entities and relationship to build the knowledge graph. Then, use the HanLP segmentation tool to generate keywords, and classify problems by CHI feature selection to improve the efficiency of QA system. By compared with other models, optimize BiLSTM model to construct the suicidal text classifier for detecting dangerous users’ speech. Finally, by computing the similarity score of template match, generate the answer. By testing the accuracy of the system, it is proved that the system can effectively answer the questions related to psychological counseling.

Key words: knowledge graph, question answering system, text classification, suicidality detection

摘要:

针对当今社会人们因缺乏心理知识而导致心理疾病患病率急剧上升的问题,构建了一个小型的心理咨询知识图谱与问答系统(Question Answering system,QA)。该系统可以帮助咨询者及时获取心理知识,也可以识别出有自杀倾向的咨询者,防止危险发生。搜集了一些心理信息数据,通过构建字典并生成实体与关系构建了知识图谱;使用了HanLP(Han Language Processing)分词工具来生成关键词,通过CHI(Chi-square)特征选择来进行问题分类,提高问答效率;通过与其他模型对比,优选BiLSTM(Bi-directional Long Short-Term Memory)模型构建了自杀倾向文本分类器来检测危险的用户发言;通过计算相似度得分来进行问题模板匹配并生成答案。最后进行系统正确率测试,证明了该系统可以有效回答心理咨询相关的问题。

关键词: 知识图谱, 问答系统, 文本分类, 自杀倾向检测