E-commerce Platform Review Text Sentiment Analysis Based On Deep Learning

With the rapid development of Internet technology,the Internet penetration rate in my country is gradually increasing,and there are more and more Internet users,and the e-commerce platform has also grown.In order to improve the user's shopping experience and collect users' real shopping opinions,all major e-commerce platforms provide comment areas to display the comments made by users who have purchased.By collecting and interpreting published comment texts for sentiment analysis,users can get the true views of other consumers on the product,and merchants can also see first-hand feedback from customers.However,with the accumulation of review texts,the massive amount of text data makes it difficult for users and businesses to easily obtain effective information,and the sentimental tendencies obtained by traditional analysis algorithms are also relatively vague.Therefore,it is very important to quickly obtain the overall emotional orientation of the review text and accurately obtain the emotional orientation of all aspects in the review.This paper is based on the improvement of basic deep learning models such as LSTM and attention mechanism.Through the study of the overall sentiment analysis of the review text and the aspect-level fine-grained sentiment analysis,obtain the overall emotional tendency of the review text and the fine-grained emotional tendency of the aspect level.For the overall sentiment analysis of the review text,the ALBERT-LSTM model for user reviews is proposed.The contextualized word vectors are obtained through the ALBERT pre-training language model,and the neural network model LSTM is combined to construct the ALBERT-LSTM model.Compare other word segmentation tools during preprocessing,and choose the fastest and best jieba word segmentation tool.ALBERT contains contextual semantic information due to pre-training on a large number of corpora,and at the same time reduces the training time with less loss,making this model have better performance than other word vector models.Combining the LSTM neural network,which is superior to the cyclic neural network that is easy to disappear and is difficult to capture long-distance effects,improves the effect of sentiment analysis.In terms of fine-grained sentiment analysis,the Bi LSTM fine-grained sentiment analysis model combined with the attention mechanism is proposed,and the effectiveness of this model is verified by experiments on the Sem Eval2014 data set.The Bi LSTM model combined with the attention mechanism uses the BERT model as the input embedding layer,and combines the advantages of the attention mechanism and the Bi LSTM network to effectively extract feature information that is valuable for the prediction of sentiment polarity in a given aspect,and improves the performance of fine-grained sentiment analysis.