基于知识图谱的问答系统设计  

Design of knowledge graph based Q&A system design

作　　者：秦丽娟[1] Qin Lijuan(Jiangsu Second Normal University,Nanjing 210013,China)

机构地区：[1]江苏第二师范学院,江苏南京210013

出　　处：《无线互联科技》2023年第18期46-48,53,共4页Wireless Internet Technology

基　　金：2019年度江苏省高校哲学社会科学研究专题项目,项目编号:2019SJB154。

摘　　要：随着网络数据的爆发式增长,知识泛滥和知识过载问题日益凸显。传统的问答系统通常采用简单的关键词匹配模式,往往无法准确理解用户的真实意图,难以提供准确的答案。为解决这一问题,文章设计了一种基于知识图谱的智能问答系统。首先,通过本体层构建、数据爬取、数据存储等步骤构建知识图谱。其次,分别采用BERT+BiLSTM+CR模型和BERT+TextCNN模型进行命名实体识别和用户意图识别。最后,使用Flask封装后台API,以便提供更加灵活和个性化的服务。With the explosive growth of online data,the problem of knowledge flooding and knowledge overload is increasingly prominent.Traditional Q&A systems usually use a simple keyword matching model,which often fails to accurately understand the real intention of users and makes it difficult to provide accurate answers.To solve this problem,the article designs an intelligent Q&A system based on knowledge graphs.Firstly,the knowledge graph is constructed through the steps of ontology layer construction,data crawling and data storage.Secondly,BERT+BiLSTM+CR model and BERT+TextCNN model are used for named entity recognition and user intention recognition,respectively.Finally,the backend API is wrapped using Flask in order to provide more flexible and personalized services.

关 键 词：知识图谱 问答系统 用户意图 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]