Research on Multi channel Text Classification Method Based on Neural Network

To solve the problem that a single neural network cannot obtain accurate text information, this paper proposes a text classification model based on neural network multi channel. Firstly, this method utilizes an ALBERT pre trained language model to obtain text vector representations, use CNN to obtain local features of the text, use GCN and BiLSTM to obtain global text feature information in both spatial and temporal dimensions, respectively. Then, this method fuse the features obtained from multiple channels. Finally, this method use softmax to calculate the probability of text category. To verify the performance of our method, this study conducted comparative experiments on the THUCNews dataset. The experimental results show that compared with a single feature extraction method, the performance of the multi-channel text classification model proposed in this paper is significantly improved. Compared with other combined feature extraction methods, the method proposed in this paper can extract richer text information. The method proposed in this paper is compared and analyzed with popular text classification models, and outperforms other models in accuracy, recall, and other indicators, further proving the superiority of the proposed method.