Research on Dual-Channel News Classification Method Combined with TextCNN and BiGRU

The language classification model based on deep learning is insufficient for extracting text features and focuses primarily on semantic information. This limitation does not address issues related to deep bidirectional representation or capturing significant text features. To address these issues, this paper proposes a dual-channel news classification method that combines TextCNN. First, the model uses the BERT pre-trained language model for text representation, converting text into dynamic word vectors. Since TextCNN has difficulty capturing the global features of short text corpora, this paper introduces a two-channel classification model that integrates TextCNN with BiGRU-ATT. Experimental results show that, on the THUC dataset, the accuracy (Acc) of the proposed model reaches $\mathbf{9 4. 1 8 \%}$.