Chinese Long Text Classification Based On Gate-Temporal Convolutional Neural Network

Le résumé fourni par la source

Convolutional neural networks can effectively extract text features but cannot adequately capture the contextual semantic information between texts, and are weak in processing long texts. In this paper, we propose a text classification model based on Temporal Convolutional Network (TCN) to address this shortcoming, and we propose Gated Temporal Convolutional Neural Network (Gate-TCN) to address the problem that the performance of the model degrades due to the introduction of noise when the number of layers of the network is deep. The model introduces the gating mechanism in the TCN module, which can effectively filter the noise and retain the effective information to improve the overall performance of the model. The Gate-TCN is used for comparison experiments on a publicly available long text classification dataset. The experiments demonstrate that compared with the traditional baseline model, the algorithm proposed in this paper has a substantial improvement in both the accuracy and F1-score. Compared with the temporal convolutional neural network (TCN), the accuracy of the Gate-TCN model is improved by 2.66% and the F1- score improved by 1.32%. The experiments show that the Gate-TCN model proposed in this paper can effectively improve the long text classification and outperforms the commonly used models.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.