基于文本匹配的电商对话系统设计

Design of E-business Dialogue System Based on Text Matching

文本匹配是自然语言处理中解决自动对话问题的关键技术,本文设计了基于表示的文本匹配模型(DSSM模型),将带有记忆功能的循环神经网络(LSTM)引入模型之中,使得模型具有更好的泛化性能。使用百度开源数据集训练出来的网络,文本匹配程度达到76.6%。将训练好的网络应用到电商对话系统中性能表现良好,具有一定的鲁棒性能和实际意义。

Text matching is a key technology to solve the problem of automatic dialogue in natural language processing.In this paper,an expression-based text matching model(DSSM model)was designed.And in order to achieve better generalization performance,the circular neural network(LSTM)memory function was introduced into the model.With the network trained from baidu′s open source dataset,the text matching degree has reached 76.6%,and the E-business dialogue system has a good performance.It is proved that the system has certain robust performance and practical significance.