Convolutional neural networks news classification based on gating mechanisms

Aiming at the problem that a single convolutional neural network often easily ignores the semantic information of the context. A convolutional neural network based on a gating mechanism (gated Linear Unit Convolutional Neural Networks) is proposed and applied for news classification. GLU-CNN model introduces a gating mechanism after the convolution layer and controls the feature information after the convolutional layer through the gating mechanism, which can reduce the gradient dispersion and also retain the nonlinear ability of the model. The experimental results show that compared with traditional convolutional neural networks, the number of classifications of different news types has increased in the same news dataset. The gating mechanism of convolutional neural networks has a certain improvement in classification effect. The model's accuracy in news classification reached 95.97%.