关键词: 混合神经网络, 特征融合, 注意力机制, 双向门循环单元, 胶囊网络

Abstract:

Most text sentiment analysis methods cannot effectively reflect the importance of different words in a text sequence and can not obtain enough text information. The text sentiment analysis model of the two-channel hybrid neural network is proposed. Firstly, the hybrid neural network layer adopts the attention mechanism respectively after the capsule network model and the Bidirectional Gated Recurrent Unit（BiGRU） model, so that it can perceive context information adaptively and extract the text features that affect text sentiment analysis. The model fuses the features extracted by the two models. Then, the results obtained by mixing the two different word vectors through the hybrid neural network layer are further fused and classified by using the Softmax classifier. The results on the three datasets demonstrate the validity of the model.

Key words: hybrid neural network, feature fusion, attention mechanism, Bidirectional Gated Recurrent Unit（BiGRU）, capsule network