Sentiment Analysis Of E-commerce Product Reviews

With the development of e-commerce,reviews on e-commerce platforms have become an important source of information to understand customer attitudes.These reviews contain user feedback on product quality.Therefore,sentiment analysis of user reviews is the basis for merchants to improve product quality and services.Sentiment classification is an important research direction of natural language processing,which has important research value and practical application value.The focus is to divide comments into positive and negative categories according to the polarity of emotion.Due to its excellent performance,emotion classification methods based on deep learning have gradually become mainstream.The sentiment classification of e-commerce reviews can help merchants and e-commerce platforms understand customer preferences and needs to improve service quality and customer satisfaction.This article is based on the sentiment analysis of the review data in the mobile phone field under the e-commerce platform,the TFIDF algorithm is improved in the keyword extraction,and combined with the deep neural network for sentiment classification.The main work is as follows:1.New words in the field of electronic products are frequently used and tend to be networked.The basic dictionary contains limited emotional words.In response to this problem,this article extracts keywords with high frequency in the field of electronic products to construct a keyword database,and selects from the text collection for high-frequency emotional words,select representative positive and negative words from the basic dictionary as seed words,and use the SO-PMI algorithm to calculate the point mutual information difference between the words that have not judged their emotional tendency and the seed words,according to the value to determine whether to expand into the keyword library.2.A training-based improvement method is proposed to obtain new weights through training enhanced texts and constrained texts,combined with the weights of keyword database training,optimize the extraction of keywords from the original TFIDF algorithm,modify the weights,and combine the Word2 vec model to form weights Word vector to further improve the accuracy of text classification.3.The improved weighted word vector is used as the input value of the CNN-LSTM hybrid neural network model,and the emotion classification model combined with the deeneural network is constructed,and the experiment is compared with the existing model.Experimental results show that the model proposed in this paper has high performance in text classification.4.From the perspective of the percentage of the training set,the accuracy of the model and the loss of the model,the model in this paper is analyzed experimentally.The results show that the percentage of the training set has a certain impact on the classification results;the model in this paper has higher accuracy,lower model training loss,and overall better classification results.