基于知识图谱和模型融合的医疗问答系统的构建

Development of medical question-answering system based on knowledge graph and model integration

基于语义解析和规则匹配融合的模型,利用少量的语义训练语料,以中文医疗知识图谱为知识基础,构建中文医疗问答系统,解决医疗领域中文语料缺乏且标注难度大的问题。该系统由语义解析模块(SPM)和答案查询模块(AQM)组成。其中,SPM由意图识别和命名实体识别组成,它们分别以BERT-Text CNN和BiLSTM-CRF模型为基础,融合了规则校正,其准确率较非融合模型分别提升9%和11%。实验结果表明,该系统回答准确率达到82%,具有较强的问题解答能力和一定的实用价值。

A Chinese medical question-answering system consisting of a semantic parsing module(SPM) and an answering query module(AQM) was developed based on the semantic analysis and rule-matched model integration using a small amount of semantic training corpuses with Chinese medical graph as its basis of knowledge to solve the insufficient Chinese corpuses and overcome the difficulty in their annotation. The SPM is consisted of intent recognition(IR) and named entity recognition(NER). The rule correction is integrated in IR and NER based on the BERT-TextCNN model and BiLSTM-CRF model respectively. The accuracy of integrated model and non-integrated model increased 9% and 11% respectively. The accuracy of the developed question-answering system in this study reached 82%, indicating that the system is can effectively solve the problems with a certain practical value.