基于词向量模型的对话短文本意图分类方法.PDFVIP

基于词向量模型的对话短文本意图分类方 法 高廷丽, 陶建华, 杨明浩, 孙梦伊, 张大伟 中国科学院自动化研究所模式识别国家重点实验室，北京100190 摘 要：人机对话系统中，在不限定用户输入语句或者不限定语句模式的情况下，环境噪声以及方言口音 语音识别错误等原因，使得计算机对用户对话意图判断错误，导致计算机针对用户的反馈语句出现“答非 所问”的情况，影响了人机对话系统用户的体验感。为了提高对话系统中意图识别的准确率，本文将词向 量的语义挖掘能力应用到对话系统的意图分类中。实验结果表明，该方法比传统的基于LDA 模型的方法和、 基于n-gram 模型的方法在短文本分类方面具有更高的准确率，在多类意图分类中，具有更好的稳定性和可 扩展性，能很好的适用于对话系统中的短文本类型的用户意图判断。 关键词 词向量，LDA，意图分类，短文本 投稿会议 CHCI2015 The method of short text intention classification based on word vector model Tingli Gao, Jianhua Tao, Minghao Yang, Mengyi Sun, Dawei Zhang National Laboratory of Pattern Recognition (NLPR), Institute of Automation, Chinese Academy of Sciences, Beijing, China Abstract In human machine dialog system, In the case of the user’s input statement (or statement mode) is not limited, due to the error of speech recognition and other reasons, will cause the system is difficult to accurately determine the users dialogue intention. It’s will cause the system appear give an irrelevant answer for users feedback statements. This situation brings the interruption of the dialogue, the user’s experience of the man-machine dialogue system is affected to a certain extent. In order to improve the accuracy of intention recognition in the dialogue system, In this paper, the semantic mining ability of word vector is applied to the intention classification of the dialogue system. Experimental results show that, the proposed method has higher accuracy than the traditional LDA based method and the n-gram based method. In the classification of the intention of more classes, the stability and scalability of this method is better. Key words word2vec, LDA, Intention classification, Short text