引入注意力机制的BiGRU-CNN情感分类模型

BiGRU-CNN Sentiment Classification Model with Attention Mechanism

自然语言处理中采用提取用户评论中的情感信息来进行情感分类的研究,但在现有的方法中很多只考虑了评论中的文本信息,而忽略了用户特征和产品特征在决定评论的情感时发挥的作用.因此,为了解决上述问题,本文提出一种引入注意力机制(Attention Mechanism,AM),将双向门控递归单元(Bidirectional Gated Recurrent Unit,BiGRU)和卷积神经网络(Convolutional Neural Netw ork,CNN)相联合的文本情感分类模型.本模型首先利用BiGRU进行序列化学习,然后引入注意力机制将较大的权重赋予给相关的特征,之后将其作为CNN的输入来提取评论文本中的特征,从而充分结合了文本的时间序列信息和局部上下文信息,将最大池应用于连接向量,分别得到用户评论文本与产品文本的特征表示,并将两者进行结合得到最终的文本特征进行分类.经实验结果表明,该模型对评论文本情感分类的作用是很大的.

Natural language processing applies the extraction of sentimental information from user comments to conduct sentiment classification research,but many existing methods only consider text information in review s,and ignore the role of user characteristics and product characteristics in determining the sentiment of review s.Therefore,in order to solve the above problems,this paper proposes the introduction of an Attention Mechanism(AM),which combines a Bidirectional Gated Recurrent Unit(BiGRU)and a Convolutional Neural Network(CNN)combined text sentiment classification model.The model first uses BiGRU for serialized learning,then introduces an attention mechanism to give greater weight to related features,and then it as the input of CNN to extract features in the comment text,thus fully combining the text Time series information and local context information,the maximum pool is applied to the connection vector,and the feature representations of user review text and product text are obtained respectively,and the two are combined to obtain the final text features for classification.The experimental results show that the model has a great effect on the sentiment classification of review text.