融合注意力机制和BiLSTM的人机交互语言训练系统研究

Research on human-computer interaction language training system integrating attention mechanism and BiLSTM

为提高英语语言训练的准确性,提出一种基于注意力机制和BiLSTM的对话意图识别方法及系统。首先,构建本人机交互对话语言训练系统框架;然后重点对系统中的自然语言理解模块和对话管理模块进行设计。其中,搭建基于融合注意力机制的语言理解模型,从而提高对语言意图识别的能力,提出基于槽框架和状态机的对话管理方法,以提高系统对话的灵活性和可扩展性;最后对构建的语言理解模型、对话管理方法及人机交互的语言训练系统进行测试。结果表明:基于融合注意力机制的英语课程训练语言理解模型对语义信息获取的准确率达到了96.1%;基于槽框架和状态机的对话管理方法适应性强、维护简单、可扩展性好;基于人机交互的语言训练系统所有功能模块都能正常运行,能满足英语自主训练的要求。

To improve the accuracy of English language training,a dialogue intention recognition method and system based on attention mechanism and BiLSTM are proposed.Firstly,construct a framework for personal computer interactive dialogue language training system;Then focus on designing the natural language understanding module and dialogue management module in the system.Among them,building a language understanding model based on the fusion attention mechanism to improve the ability to recognize language intentions,and proposing a dialogue management method based on slot framework and state machine to improve the flexibility and scalability of system dialogue;Finally,the language understanding model,dialogue management method,and human-machine interaction language training system constructed were tested.The results show that the accuracy of semantic information acquisition in the English course training language understanding model based on the fusion attention mechanism reaches 96.1%;The dialogue management method based on slot framework and state machine has strong adaptability,simple maintenance,and good scalability;All functional modules of the human-computer interaction based language training system can operate normally and meet the requirements of English autonomous training.