Sentiment Analysis Of User Reviews Based On Dilated Convolution

With the development of Internet technology and the promotion of e-commerce platforms,more and more users choose online shopping and online consumption.Therefore,sentiment analysis of consumer evaluation of goods or services has become a current research hotspot.Under the background that deep learning is widely used,as the standard network in the image field-the convolutional neural network,because it cannot handle long sequences of text and the problem of information loss in the field of natural language processing,the convolutional neural network is in the text The field of sentiment analysis is always used in conjunction with sequential networks.In this paper,from the two ideas of improving the structure of convolutional neural network and the composition of text sentiment,two user sentiment analysis models-SP-DCNN and Tree-DCNN are proposed.The core work of this article includes two aspects:First,in order to improve the convolutional neural network so that it can handle the user comment sentiment analysis task,this paper starts from improving the structure of the convolutional neural network and designs a dilated convolutional neural network based on the cascade segmentation structure-SP-DCNN.The model uses a dilated convolution to replace the pooling layer to solve the information loss problem of the convolutional neural network.At the same time,it uses a cascade connection method and combines the low-order features and high-order features extracted by the model for the final emotion.Classification prediction to solve the problem that convolutional neural networks cannot handle long sequences.In addition,SP-DCNN also adds emotional word vectors and multi-core feature extraction,so that the model can extract the emotional features of word groups in the text.Second,from the perspective of the tree structure,combined with the hierarchical relationship between the textual sentiment,sentence sentiment,and word sentiment of user reviews,this paper also designs another new user review sentiment analysis model—TreeDCNN.The core structure of the model is still a dilated convolutional network,and its focus changes from the overall text sentiment to from the sentiment of words to the sentiment of highorder sentences to the sentiment of the final chapter.Therefore,the model uses a hierarchical network structure,which has a certain improvement in model effect and interpretability.This article not only theoretically proposed two models of user comment sentiment analysis,but also conducted experiments on the Chn Senti Corp open source data set separately to achieve the desired effect.