Text Classification Method Based on BiGRU-Attention and CNN Hybrid Model

Abstract

Aiming at the problem that traditional Gated Recurrent Unit (GRU) and Convolution Neural Network (CNN) can not reflect the importance of each word in the text when extracting features, a text classification method based on BiGRU Attention and CNN is proposed. Firstly, CNN was used to extract the local information of the text, and then the full-text semantics was integrated. Secondly, BiGRU was used to extract the context features of the text, and attention mechanism was used after BiGRU to extract the attention score of the output information. Finally, the output of BiGRU attention was fused with the output of CNN to realize the effective extraction of text features and focused on the important content words. Experimental results on three public datasets showed that the proposed model was better than GRU, CNN and other models, which can effectively improve the effect of text classification.