Design Of User Review Supervision System Based On Deep Learning

This paper through using Text Classification Algorithm of Machine Learning proposed a User Comment Monitoring System based on Deep Leaning.The main-work contents are as follows:1)Use BERT feature extraction to generate context-dependent word vectors.Instead of using the traditional Word2 Vec word vector acquisition method,this method uses the Self-Attention in the two-way deep Transformer model to train the word vector.The BERT model can enable the vocabulary to see the context information during training,and then the word vector can merge the context-related characteristics.Therefore,the obtained word vector not only contains the context information,but also has different representations in different sentences,as a result,the expression becomes more accurate.Get rid of the defects of traditional word vector fixing representation.2)A convolutional neural network text classification algorithm based on BERT feature extraction is proposed.On the basis of the above-mentioned word vector method,the word vectors are spliced through the output of the model,and the spliced word vector matrix is used as the input of the convolutional neural network.The characteristics of the convolutional neural network are used to further extract the word vector features,so that the accuracy of text classification will be effectively improved.3)The supervision system of forum comments is developed.The system is mainly divided into three modules,which are comment detection module,user function module and administrator management module.The comment detection module is responsible for detecting users’ toxic comments.The users’ function module realizes some basic activities of users while going through the forum.The administrator management module is mainly responsible for the user’s accessing rights.Finally,this article combines text classification algorithms with practical applications to build a user comment monitoring system based on deep learning.