Research And Analysis Of Commodity Reviews Based On Text Mining

With the constant change of Internet speed,more and more people are eager to buy necessities on the platform of e-commerce.And for any kind of goods,the first attention of consumers is the product reviews,to determine whether the product is worth buying.While the product reviews represent the most authentic consumer evaluations of the products,the textual information of these evaluations contains great value.Enterprises can quickly focus on the advantages and disadvantages of the products from the reviews,consumers shopping habits and some ancillary services,such as some of the shortcomings of the corresponding transformation management,enhance consumer satisfaction and expand the influence of enterprises.However,the traditional social research cannot meet the requirements of today's fast-developing society.What we need more is a way to get the information we want from the mass comments quickly,therefore,based on text mining,the analysis of comment emotion comes into being as the Times require.In this paper,a brand of computer reviews under the Jd.com is analyzed,and LDA subject model is used to extract the subject words of good reviews and bad reviews,then sentiment classification is realized by sentiment dictionary and Decision Tree respectively.After the analysis of the brand computer reviews is completed,consumers can be provided relevant choice of recommendations.First of all,after the data preprocessing,we draw the word cloud map and find out that for the consumers,the brand of computer appearance and speed are the most important.After that,the number of subjects is optimized by using cosine similarity,and then the top ten words in each subject are output by using LDA subject model,and then the word cloud is drawn.Whether it is good or bad,are mainly based on computer performance and after-sales service to evaluate the two themes,then businesses in the future can be improved in these two areas.Finally,it is beneficial to use two methods for text sentiment classification,one is sentiment classification based on sentiment dictionary and the other is sentiment classification based on decision tree,which makes the classification effect of the model better.Through the precision rate,recall rate,F1 value of these indicators,we can see that in the data set of this paper,the classification method based on emotional dictionary performs better.It also shows that these two methods can be used in emotion classification.