Chinese text sentiment classification method based on dual-channel attention capsule network

Abstract

In recent years, text sentiment analysis has received close attention from researchers. The CNN-based neural network is frequently used in this field. Capsule network can overcome the problem of information loss caused by pooling layer in CNN-based neural network, but the capsule network cannot selectively focus on important words in the text. To solve this problem, this paper proposes a text sentiment analysis method based on the attention mechanism of the capsule network. First, use the cw2vec model and the SAT model to train word vectors and use them as the input of the two channels of the method, then combine the attention mechanism and the convolutional neural network to extract the characteristics of the two channels, and then input the capsule network to realize the emotion classification. The static routing mechanism is also used to reduce the complexity of the model, and it has a higher accuracy than the dynamic routing mechanism. We verified the effectiveness of the model on the Tan Songbo hotel review data set and we further crawls a large amount of public opinion data related to Meteorology and early warning from Weibo to form different fields. the model is also better than many sentiment analysis methods.