Topic Mining And Sentiment Analysis Of User Reviews On Audio Book Application

With the rapid development of Internet technology,people's reading habits are gradually becoming digital and fragmented.Audiobook application is favored by users because of its rich content and strong companionship.In recent years,audiobook apps in app stores have emerged one after another,mobile reading and mobile music apps have also begun to have built-in listening services.However,due to the serious homogeneity of content in audiobook apps,competition among similar APPs is fierce;user transfer costs are low,and loyalty to the platform has not yet been established.Therefore,the research on the opinions of audiobook app users is particularly important.The mobile application store contains a large number of user reviews,which are the true expressions of users' opinions after using products or services.Topic mining and sentiment analysis of user reviews can discover the main aspects of the user's attention to the APP,as well as the emotional expression of each theme.Operators can based on the research conclusions,targeted improvements to product deficiencies,provide users with better services,and enhance platform competitiveness.This research selects reviews of 5 representative audiobook apps in mainstream application stores,and uses text mining technology to conduct topic mining and sentiment analysis on a total of more than 80,000 user reviews.Firstly,based on TFIDF,extract the subject terms in user reviews,and find that the topics that users pay attention to involve multiple aspects,among which there are more opinions about advertisements and members;then based on the LDA topic model,the user review topics are excavated,and the user reviews are found.the themes can be summarized into seven themes of audio content,APP function,APP performance,paid service,anchor,customer service,and interface interaction.After the topic dimension is determined,the topic sentence in the review text is identified based on the machine learning method.According to statistical analysis,users' attention to the seven topics is ranked as follows: audio content,interface interaction,APP performance,paid service,APP function,customer service,and anchor.Using machine learning methods to perform sentiment analysis on the comments under each theme,it is found that the emotional performance of the three themes of audio content,anchor,and APP function is better;the emotional performance of the four themes of APP performance,paid service,customer service and interface interaction Poor,and the emotional performance of the interface interaction theme is the worst.The main reason is that users have more negative comments on the advertising interface.Finally,combined with the above research conclusions,five aspects of improvement countermeasures are provided for audiobook APP operators,so as to provide a reference for operators to better serve users and enhance product competitiveness.