Sentiment Analysis Of Multi-domain Product Reviews Based On Deep Learning

With the decline in the cost of Internet technology and the increasing popularity of smart handheld mobile devices in recent years,more and more people are beginning to participate in the Internet,shopping through the Internet,browsing some international news,watching sports live broadcasts,or Some platforms participate in topic discussions.Among them,the increasing online shopping behavior has stimulated the rapid rise of domestic e-commerce platforms,such as:Taobao,DangDang,JingDong,Suning Tesco and so on.The expansion of these platforms not only provides a large amount of commodity review data,but also enriches the fields involved in the data,from the initial basic necessities of life to all the daily necessities in people's lives.Therefore,by performing sentiment analysis on a large amount of subjective data,the macro emotions or attitudes of the crowd to certain entities can be obtained,and the results can be applied to other tasks,such as product recommendation system.Moreover,in recent years,artificial intelligence and deep learning technologies have gradually become hot topics,and deep learning technology has been successfully applied to many fields,including:biomedicine,pattern recognition,and natural language processing.On most tasks,deep learning has outperformed traditional machine learning algorithms in terms of performance and model complexity.Therefore,this paper uses deep learning to conduct sentiment analysis on multi-domain e-commerce commodity review data.In the representation of the text,the BERT language model is used for pre-training to obtain the vectorized representation of the comment data.In the selection of the model,this paper proposes a sentiment analysis model based on capsule neural network,and realizes the emotion based on the cyclic neural network.The analysis model and the sentiment analysis model based on convolutional neural network are used as comparison to verify the validity of the proposed model.At the same time,the support vector machine is used as the representative of traditional machine learning and compared with the deep learning model.The experimental results show that the sentiment analysis model based on capsule neural network is superior to the other two deep learning models in the performance of sentiment classification.At the same time,the experiment proves that the deep learning model performs better than the support vector machine in the sentiment analysis.