User Review Satisfaction Based On Text Mining

With the overall enhancement of national cultural confidence and national confidence,the reactivation of traditional culture causes "national hot tide".Under the "national hot tide",foreign international brands have not stopped their pace of progress.The severe market competition makes domestic sports brands have to think again about how to "stand firm" in the domestic market and improve their competitiveness.Online comment is a lowcost channel to obtain user feedback.Through the comment content,users can obtain the emotional tendency of various attributes of goods,which can provide reference for consumers to shop and help merchants to improve product quality.So this article selected the representative of domestic sports brand Li Ning and international sports brand Skechers,climb took the Jingdong mall and Taobao platform online review of data,using the method of text analysis,the Li Ning brand and emotional Skechers brand shoes reviews are classified,and in-depth digging,the theme of the theme for each brand to satisfaction analysis,And put forward suggestions to Li Ning brand,has certain practical significance.This paper is divided into five parts: The first part explains the research background,and expounds the current research direction and analysis methods of topic model,emotion classification and satisfaction analysis by domestic and foreign scholars.In the second part,the text segmentation technology,word quantization technology,topic model and the principle and structure of TextCNN and LSTM neural network are described in detail,which makes theoretical preparation for the construction of the following models.The third part explains the sources of online comment data sets of footwear products and makes a basic descriptive analysis of the data sets of Jingdong and Taobao platforms respectively.Use oversampment method and EDA to deal with unbalanced data set,combined with TextCNN and LSTM neural network method to build classification model under different text enhancement strategies respectively,give emotional classification of Jingdong review text.According to the evaluation index,the optimal model is selected to classify and predict the unlabeled comment text on Taobao platform.The fourth part uses GSDMM theme model to dig the theme of Shoes of Li Ning brand and Skechers brand,and get the emotional score of different themes;According to the characteristics of theme features and online shopping,the sub-themes of the same category are summarized,and the two brands are analyzed from the perspective of comprehensive satisfaction and parent satisfaction.The fifth part is the conclusion and suggestions,according to the previous several parts of the theory and empirical analysis,li Ning brand suggestions.According to the research findings,firstly,consumers generally pay attention to the appearance level,size and quality of workmanship of footwear products.Second,aiming at the unbalanced data set in this paper,it is concluded that TextCNN model under EDA method has the best classification effect.Third,based on the analysis of theme satisfaction,the score of Li-Ning shoes in the quality of workmanship,appearance design,price,size,logistics and national tide effect and other themes are higher than 0.9;Skechers shoes scored less than 0.8 on the subject of merchandise logistics and customer service.Then,from the product quality,product price,customer service three aspects of li Ning brand recommendations.Finally,the future prospect of data set selection and model optimization is proposed.