Design And Implemention Of APP Store Review Analysis System Based On Sentiment Analysis

With the rapid development of mobile Internet and the popularization of smart phones,it has become people's daily life to access the Internet and consume Internet services by using smart phones.As the most convenient and fastest way for users to access a mobile application(APP),APP store is a significant part of the brand ecology establishment for the phone manufacturers.Reviews generated by users after they experience the APPs will contain opinions and sentiment information,which are of great value for both APP developers and manufactures to mine and analyze.Based on the structural features of an APP store,user reviews can be divided into two types:APP evaluation reviews and problem feedback reviews.Employing sentiment analysis technology can obtain users' true evaluation of an APP,or extract problem aspects and users' opinions in the feedbacks as well.Meanwhile,APP stores are also faced with troubles of spam or illegal reviews.Therefore,it's necessary to detect and filter the spams in the raw review data first.This thesis mainly studies the employment of sentiment analysis methods with different granularities to analyze the APP evaluation reviews and problem feedback reviews respectively in an APP store.In this thesis,a spam filtering model,an APP evaluation analysis model and a feedback problem analysis model are designed.On this basis,an APP store review analysis system is designed and implemented.Main researches of this thesis are as follow:(1)Researches on spam review detection and filter of the raw review data in an APP store.A spam review filter model for the APP store reviews based on filter lexicon and support vector machine is designed.(2)Researches on setting up a sentiment lexicon set to classify users' sentiment polarity of an APP.The idea of Sentiment-oriented Pointwise Mutual Information(SO-PMI)algorithm is applied to expand the domain lexicon.An APP sentiment score rating model based on sentiment lexicon set is designed.(3)Researches on analyzing feedback reviews on a fine-grained level.An existing target extracting algorithm called Double Propagation(DP)is applied to extract the problem aspects and sentiment expressions in the feedback reviews.Considering that the part of speech of the target words can be verbs,the DP algorithm is improved by adding 4 new rules for extraction.A problem feedback review analysis model based on improved DP algorithm is designed.(4)Based on the three review analysis models proposed above,an APP store review analysis system is designed and implemented.