Research And Application Of Key Technology Of Sentiment Analysis Based On Mobile Game Comment

With the development of mobile network and the widespread use of mobile devices,mobile games have become a way for people to enjoy themselves in their spare time.People can download mobile games at many app markets and write comment about these games on markets.For game operations,they can observe the players' response and reputation on the game activities and new version of game from their comments,so game operations can make the games better,which can take advantage in the competition.However,with the passage of time,comments will increase.It is difficult to analysis the massive comments by manual work.Therefore,it is necessary to build sentiment analysis system to analyze these comments.This paper studies the key technology of sentiment analysis based on domain of mobile game including clustering the same kind of games,collecting of unknown word to build lexicon in domain of mobile game and analyze the sentiment of comment and extraction of feature and sentiment word in comment.Based on these key technology,I have built a sentiment analysis system on domain of mobile game using Java and some tools,such as Hadoop?Spark?Kafka?Elasticsearch?Zookeeper?MySQL?Spring MVC and etc.The main tasks of this paper include:(1)The Crawling of comments and basic information of games,clustering the same kind of games and collecting unknown word in game's corpus.The crawler can crawl comments from many sources using Spark,Kafka,Redis and etc.(2)Sentiment Analysis based on games' comments.This paper studies how to use Co-Training and Pu-Learning to label the polarity of large-scale comments.This paper also train NB-LR model using liblinear to predict the polarity of comments.This paper extract the feature and sentiment word using Dependency Parser and fix bad base by training a SVM model.The result shows that above model can classify the comments and this method works well.(3)The design and implementation of the sentiment analysis system.Based on the above key technology,this paper creates a sentiment analysis system using Spring MVC as the web framework,Elasticsearch as the search engine,MySQL and HBase as the storage engine,Spark as the calculation framework,Thrift as RPC framework.