Research On User Product Reviews Based On Text Mining Technology

With the rising enthusiasm of the public for the consumption of domestic brands and the quiet rise of domestic brands,especially the performance of sports brands in the ranks of domestic products is particularly prominent.Taking the Hongxing Erke brand as an example,in the donation event for the rainstorm disaster area in Henan,the Hongxing Erke brand donated50 million materials to the rainstorm disaster area for the first time,which aroused the keen attention of the general public.In this context,this paper takes Hongxing Erke brand as an example,uses text mining technology to analyze the evaluation content of Hongxing Erke brand online shopping users,and further reflects the brand's image in consumers' minds through the analysis of user evaluation content,making the brand more perfect.First,select products with more than 2,000 reviews,and use Python software to crawl the review information related to the user's review content,and finally obtain more than 16,000 review information.Then,the acquired text data is cleaned,segmented,established a stop word dictionary,and removed stop words.Finally,the sentiment classification is performed on the processed data.The sentiment classification methods used in this paper mainly include sentiment dictionary,machine learning algorithm and LSTM neural network.For sentiment classification,firstly,the text data is roughly classified by the method of sentiment dictionary classification,and then the data set is accurately divided in the form of manual annotation.Then,according to the corresponding classification algorithm,the training set and the test set are respectively trained and tested,and finally the best classifiers for sentiment classification are obtained as Naive Bayes and LSTM neural network.It can be seen from the classification results of emotional polarity that the overall emotional tendency of consumers towards Hongxing Erke brand products is positive,but some consumers still have negative emotional attitudes.Therefore,next,according to the classified positive evaluation data and negative evaluation data,feature extraction is performed,a semantic network graph is constructed,and an LDA topic model is established.By constructing a semantic network model,it is concluded that the comfort of this brand of footwear is most recognized by consumers;there are still some defects in quality;logistics and transportation are the most problematic aspects of consumers.Through the analysis of the LDA topic model,it is concluded that the positive evaluation content reflects that the overall concept of the brand product is available and intact.That is,the overall sales structure of the brand product is complete,which can meet the needs of consumers to a certain extent and achieve the relative satisfaction of customers;the negative evaluation content reflects the price fluctuation,non-standard size,logistics and after-sales of the brand product.,and consumers are more inclined to buy in brick-and-mortar stores than in online stores.Finally,make a comparative analysis based on the above conclusions,make full use of the information reflected in the review data,and put forward reasonable suggestions and opinions for businesses and consumers.