融合文本内容和标签的中文商品评论情感分类

Affective Classification of Chinese Product Reviews by Combining Text Content and Labels

为了解决目前基于深度学习的情感分析方法对语言知识和标签信息利用不足的问题,本文对基于深度学习的情感分析方法加以改进,提出一种融合文本特征和标签信息的中文文本情感分析方法。该方法首先对文本进行预处理,然后分别使用句法规则和深度学习方法提取文本的特征信息。在句法规则部分,通过构建句法规则提取文本中情感倾向更加明确的信息,并将其与标签描述信息进一步融合,突出情感信息;在深度学习部分,通过结合注意力机制和门控循环神经网络提取文本的高阶特征。最后,将上面两部分提取的特征信息进行融合并使用分类器对其进行分类,得到文本的情感倾向。实验结果表明,与传统情感分析方法相比,使用本文方法进行情感分析的效果更好。

In order to solve the problem of insufficient utilization of language knowledge and label information in the emotion analysis methods based on deep learning,in this article we improve the emotion analysis method based on deep learning,and propose a Chinese text emotion analysis method that combines text features and label information.In our improved method we first preprocess the text,and then use syntax rules and deep learning methods to extract the feature information of the text respectively.In the part of syntax rules,we extract the more explicit information of emotion tendency in the text by constructing syntax rules,and combine it with the label description information to further highlight the emotion information.In the part of deep learning,we extract the high-level features of the text by combining the attention mechanism and gated recurrent neural network.Finally,we combine the features extracted from the above two parts and use the classifier to classify them to get the emotional orientation of the text.The experiment showed that the method proposed in this article was better than the traditional emotion analysis methods.