Sentiment analysis of microblogs based on dual-channel feature fusion
This paper proposes a microblog sentiment analysis model based on dual-channel feature fusion.Firstly,the dynamic word vectors obtained from the BERT pre-trained language model are used as the input of the sentiment classification model,and then a dual-channel feature extraction network is used for feature extraction,which extracts the local features of the text using TextCNN-Attention on one hand,and extracts the global thematic features of the text using the neural thematic model based on graph convolutional neural network on the other hand.Then the local and global features are fused to get the final text vector,and finally the sentiment polarity is output by Softmax.Experiments are conducted on the constructed microblog comment text dataset,and the F1 value of the model reaches 91.36％,having an improvement of 0.73％～8.82％compared with the mainstream baseline models,which verifies the effectiveness of the model on the task of sentiment analysis.
sentiment analysispre-trained language modelgraph convolutional neural networkneural thematic modelfeature fusion
万方数据
维普
