Research On Sentiment Analysis Based On Brand Mobile Phone Reviews

With the rise of the e-commerce industry and the rapid development of the Internet,online shopping accounts for a growing proportion of our lives,which not only makes it convenient for consumers to shop,but also makes people have higher pursuit for the brand quality of goods.Based on the carrier of the e-commerce platform,a comment-oriented consumption method has gradually expanded in the market.When consumers purchase goods,they often refer to relevant reviews,which may affect or adjust their buying intentions;Comments can also help merchants understand the emotional tendency of consumers and assist them in public opinion analysis,product optimization,user portrait and marketing decisions in order to increase economic efficiency.However,in the face of fast-growing data,it is not necessary to use inefficient manual processing.What is needed is to develop a tool that can automatically analyze the sentiment tendency of product reviews,and to mine valuable product attributes and sentiment trends from massive reviews.Consumers are better able to purchase goods,which also effectively helps businesses better understand consumer perspectives.This paper studies from two aspects: the coarse-grained emotion classification of mobile reviews and the fine-grained emotion analysis of mobile reviews.Coarse granularity is mainly to analyze the emotional tendency of the comment,whether it is positive or negative;Fine granularity is mainly to mine commodity attributes and consumer views from mobile reviews,and to confirm their emotional polarity and attribute types,and put it in the corresponding topic.Specific research results are as follows:(1)The coarse-grained emotion classification adopts deep learning algorithm that introduces attention mechanism.In view of the shortcomings of the existing algorithms,which have no interrelation between antecedent and antecedent pairs,and the training sequence can only be a temporal sequence,attention mechanism is introduced into the model.In this paper,an improved version of GRU and bidirectional LSTM is proposed,which overcomes the shortcomings of the original algorithm and makes a comparative analysis with the based algorithm,a better Transformer model is also used.The experimental results show that the accuracy rate of the original algorithm model LSTM is 59%,while the accuracy rate of GRU is not much different from that of the LSTM model,the accuracy of the Transformer model can reach 96%,while the accuracy rate of the improved GRU and bidirectional LSTM can reach 97%,which is much higher than that of the single model,and also proves the effectiveness of the attention mechanism.(2)The fine-grained emotion analysis adopts the topic clustering method based on word2 vec.In order to solve the existing problems of common theme clustering models,in this paper the LTP of Harbin Institute of Technology is used to preprocess the comments,and the mobile attributes most concerned by consumers were found from the comment data as the theme words.Then use word2 vec to cluster the topics and set up the topic attribute dictionary,and categorize the attributes into the corresponding topic.Next,SnowNLP is used to calculate the emotional value of each comment under each topic and divide the positive,neutral and negative emotions according to the emotional value and corresponding threshold.Finally,the emotional analysis and comparison of attributes of different brands and the emotional score of a brand are presented in a visual form,so as to achieve the purpose of comparative analysis of the same attributes of different brands and different attributes of the same brand.The experimental results show that for the same attributes of different brands,comparative analysis can be conducted to clarify the advantages and disadvantages of each brand of mobile phone,so that consumers can choose more appropriate brands.For different attributes of the same brand,the degree of attention paid to each attribute and the degree of emotion of consumers can be quantified,so as to help consumers have a more comprehensive and specific understanding of the brand and merchants can also optimize their products to bring greater benefits.