Sentiment Analysis Of Text Base On Deep Learning

With the rapid development of Internet,the transmission of information is becoming more and more quickly.People interact more frequently in the information via the Internet.How to effective access to relevant events,product,activity assessment information and dig out some valuable content,become a study focus in the Internet technology.Therefore,the Internet public opinion analysis technology arises at the historic moment.The text sentiment analysis to determine user information emotional attributes is a key technology of the public opinion analysis,which play an important role to understand public opinion tendency.Traditional sentiment analysis algorithm has some problems,such as high dimension feature,comprehensive words sequence information in astatement.we try to improve the problems existing in the traditional algorithm.In this paper,the main work includes:1.This paper improves a support vector machine(SVM)based on naive bayes fusion classification algorithm.n-gram model and neural network model was used to optimize the document feature vector,effectively improved the characteristic dimension problem in traditional method,and joined the phrase sequence information.2.This paper implements a language model based on recurrent neural network model,which can effectively obtain statements complete sequence information,thus improving the accuracy of the sentiment analysis.3.This paper mix the traditional sentiment analysis algorithm and two improved methods together,which further improves the accuracy of text sentiment analysis.