融合敏感词规则和字符级RCNN模型的用户意图识别

用户意图识别是基于用户对话用语识别用户的真实对话意图,是人机对话研究中的一项关键任务。针对现有用户意图识别方法的不足,提出融合敏感词规则和字符级RCNN模型的用户意图识别方法。构建敏感句子与敏感词词典,并通过规则及相似度匹配策略对特征明显的对话进行意图识别。针对类别特征不明显的对话提出深层语义分类模型,该模型以单字符串作为输入序列,利用RCNN模型构建意图分类框架,既可以避免分词结果不准确带来的错误传导问题,同时利用字符的分布向量表示方法还可以获取句子的深层语义信息。实验结果表明,该方法在两个数据集上都取得了较好的结果,明显优于传统的意图识别方法。

USER INTENT RECOGNITION BASED ON SENSITIVE WORD RULES AND CHARACTER LEVEL RCNN MODEL

User intention recognition is a key task in the research of human-machine dialogue based on real conversation intention of identifying users with dialogue language.Aiming at the shortcomings of existing methods of existing user intention recognition methods,this paper proposes a user intention recognition method,which integrates sensitive word rules and character-level RCNN model.The sensitive sentence and sensitive word dictionary were constructed,and intention recognition of dialogues with obvious features was identified by rules and similarity matching strategy.Then,a deep semantic classification model was proposed for the dialogue with inconspicuous category characteristics.The model used a single string as the input sequence,and used the RCNN model to construct the intent classification framework,which could avoid the error conduction problem caused by inaccurate word segmentation results.At the same time,the distribution vector representation method of the character could also be used to obtain the deep semantic information of the sentence.Experimental results show that the proposed method achieves good results on both data sets and is obviously superior to the traditional method of intent recognition.