Comparison Of Sentiment Analysis Methods For Chinese E-commerce Comment Based On Deep Learning

With the vigorous development of the Internet and the rapid development of information technology,the scale of e-commerce users in China is increasing day by day.E-commerce competition is becoming increasingly fierce.In order to broaden their own development,e-commerce platforms pay more attention to the comments left by mass customers on their platforms.These comments include all kinds of subjective feedback from users on products,services and platforms,which helps e-commerce to explore the public's interests and their own shortcomings,so emotional analysis emerges as the times require.The traditional sentiment analysis method based on emotion dictionary and traditional machine learning has poor performance and consumes a lot of manual annotation cost.In this paper,the deep learning method,combined with the pre-training model and attention mechanism,is used to classify the emotion from the review data of an e-commerce appliance.The bidirectional LSTM model and the Bert model are compared emphatically,and this paper overcomes the problem of data imbalance and achieves good classification effect,at the same time,we get a good review of the Bert model of sentiment classification on medical cosmetology.