基于深度学习的中文文本分类方法

Chinese text classification algorithm based on deep learning

为提升中文文本的分类效率,改善梯度消失、梯度爆炸以及语义信息提取不准确等问题,提出基于深度学习的中文文本分类模型ResCNN-LSTMA。对残差网络和卷积神经网络的组合方式进行研究,发现二者组合能够避免梯度消失和梯度爆炸的情况。分析注意力机制与长短期记忆网络组合的形式对文本分类的效果,二者组合能够在保留上下文语义信息的同时对语义信息进行提取,改善特征提取不全面的问题。通过使用搜狐新闻数据集训练模型,验证了该模型分类效果的准确性和有效性。

To improve the classification efficiency of Chinese text,improve the problems of gradient disappearance,gradient explosion,and inaccurate semantic information extraction,a deep learning-based Chinese text classification model ResCNN-LSTMA was proposed.The combination of residual network and convolutional neural network was studied.It was found that the combination of the two models could avoid the situation of gradient disappearance and gradient explosion.The effects of the combination of attention mechanism and long short term memory network on text classification were analyzed.It is verified that the combination of the two models can extract semantic information while retaining contextual semantic information,and improve the problem of incomplete feature extraction.By using the SogouCS data set to train the model,the accuracy and effectiveness of the classification effect of the model are verified.