Research And Implementation Of Text Classification Method Based On Convolutional Neural Network And Topic Model

With the continuous update and dissemination of Internet information,the phenomenon of information overload has become more and more serious.As text information is the basic information carrier,its explosive increase of multi-category information puts forward more intelligent requirements for text classification tasks.In response to this problem,this topic has studied the text classification method based on deep learning,and carried out detailed analysis and improvement on the convolutional neural network framework and its classification model.The details of the research are as follows.1.In view of the lack of topic features in the convolutional neural network classification model,this topic combines deep learning and topic model theory to propose a convolutional neural network classification model fused with neural topic models.The existing text classification model based on deep learning fails to fully consider the differences between documents of different topics.That is to say,the classification model fails to fully consider the feature enhancement brought by global features and topic features,which may be the key to restricting the further improvement of the accuracy of the classification model.In response to this problem,this topic uses Convolutional Neural Networks to encode text information,and introduces Prod LDA to mine potential topic information to make up for the lack of topic features.The experimental results show that the model has better classification performance than the classification model that does not fully consider the topic information,which proves that this method can improve the problem of missing topic information in the classification model.2.In order to further improve the feature extraction capability of the convolutional networks in the classification models,a feature fusion algorithm based on BERT and CNN was proposed.Firstly,BERT language model is used to replace the traditional Word2 Vec for semantic representation to capture semantic information more comprehensively.Secondly,a dynamic DK-Max Pooling method is proposed in the convolutional neural network to replace the maximum pooling operation,so as to better capture feature information except the the maximum value and reduce noise data interference.The experimental results show that the introduction of feature fusion algorithms can further improve the accuracy of the classification model.3.Designing and implement a text classification system based on deep learning.The system takes the algorithm model proposed in this subject as the core,and designs and implements core functional modules such as data labeling,classification model evaluation,and classification model prediction for the classification problems.