Research On Influencing Factors Of User Satisfaction Based On Text Analysis Of Online Comments

In recent years,the Internet e-commerce industry has developed rapidly,and earning traffic dividends cannot be the main way to make profits.In order to play a role in the fierce market competition,the platform is constantly looking for new innovative models that can adapt to changing consumer needs and consumer preferences.After the basic needs are met,consumers are paying more and more attention to the service experience during the consumption process,and the service experience has a greater impact on consumer loyalty and consumer psychology.Therefore,the e-commerce platform is also continuously upgrading its products and services.The effectiveness of the service experience is mainly measured by user satisfaction.User satisfaction is achieved by consumers scoring and commenting on products or services,which avoids the omission of information from the traditional artificial design of questionnaires,and effectively reduces costs.Therefore,it is extremely necessary to extract effective information from the review text for user satisfaction research.With the widespread application of machine learning technology,a series of studies using text mining methods to analyze comment texts have appeared,most of the research on user satisfaction of review texts is to find the user's attention dimension.At present,there are few relevant studies on specific analysis of different dimensions,and more conclusions are drawn on overall satisfaction.In view of the above-mentioned lack of research,this paper uses text mining methods to improve various influencing factors of user satisfaction,and build a complete user satisfaction index system.This article takes JD's self-operated mobile phone review text as the research object,and based on natural language processing technology,focuses on the user satisfaction of user experience to model the mobile phone review text.The main research contents are as follows:First,in order to improve the feasibility of data modeling,preprocessing text data mainly includes text deduplication and text segmentation.Text deduplication can eliminate redundant text and increase the effectiveness of comments.Text segmentation is the basis for data modeling.Secondly,to extract the influencing factors of user satisfaction,it is necessary to perform keyword extraction to improve the accuracy.The keyword extraction method selects the TextRank algorithm,which is suitable for the review text corpus.On this basis,the keyword K-means clustering is used to obtain the influencing factors of user satisfaction,which are eight categories:price,appearance,logistics,technology,performance,brand,after-sales,and gifts.Thirdly,empirical analysis of the factors influencing user satisfaction to build a framework of user satisfaction index system,including the sentiment value and index weight of the index.By building a neural network LSTM model to calculate the emotion value,the overall user satisfaction and the satisfaction of various influencing factors are obtained.The calculation of the weight of the user satisfaction index takes into account two factors,one is the importance of each influencing factor,and the other is the consistency between the emotion expressed by the influencing factors and the overall emotion.The factors that affect user satisfaction are as follows:technology>performance>price>brand>logistics>appearance>gifts>after sales.And compare the satisfaction of different mobile phone brands,it has been concluded that domestic mobile phones user satisfaction is more satisfied than overseas mobile phones user satisfaction.Huawei mobile phones have the highest overall satisfaction,while Apple mobile phones have the lowest satisfaction.Finally,based on the results of empirical research,this article makes substantive suggestions for mobile phone brands and consumers.This article has three innovations:(1)This article uses comment text to update the user satisfaction index system,and we reclassified and supplemented the factors that affect user satisfaction;(2)Before using the neural network LSTM model to emotion analysis,text data annotation was performed through SnowNLP,which improved classification accuracy and avoided the problem of inconsistent scoring and comments when using data annotation for scoring;(3)Two factors were considered in calculating the weight of the factors affecting user satisfaction.The first is the importance of influential factors,measured by the ratio of the number of comments included in each influencing factor to the total number of comments.The second is the consistency of the emotional tendencies of each influential factor with the overall emotional tendencies.Increasing the sentiment trends from three categories to five categories helps refine the sentiment tendency classification and improve the accuracy of emotional consistency scores.