摘要
针对传统的问答系统普遍存在回答准确率不高、语义识别能力差等问题,提出一种结合双向长短时记忆网络(Bi-LSTM)和注意力(Attention)模型的问答系统。利用生成的句向量,学习句子中的语义特征以及问答之间的匹配关系,获取上下文信息;融合注意力模型,能够找到对话的主题信息,从而为用户做出精准的回答。实验结果表明,该系统的回答准确率高于其他模型,可达到80.76%。
Traditional question and answer systems generally have problems such as low accuracy of answering and poor semantic recognition.In order to solve the above shortcomings,this paper proposes a question and answer system based on Bi-LSTM and attention model.By using the generated sentence vector,the system learned the semantic features of sentences and the matching relationship between questions and answers to get the context information.It integrated the attention model to find the topic information of the dialogue,so as to make accurate answers for the users.The experimental results show that the answer accuracy of our question and answer system is higher than other models,which can reach 80.76%.
作者
邵曦
陈明
Shao Xi;Chen Ming(College of Telecommunications and Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210003,Jiangsu,China)
出处
《计算机应用与软件》
北大核心
2020年第10期52-56,共5页
Computer Applications and Software
基金
国家自然科学基金面上项目(61872199,61872424)。
