Analysis Of Chinese Paragraphs Emotion Based On Naive Bayes

With the rapid popularization and development of WEB2.0, the microblogging, review sites, stick and other network community increasinglydeveloped, people express personal perspective on things, events andcommodity through these channels then appears a large number of text withpersonal feelings, attitudes or opinions, and these text sentiment analyses withthe larger commercial and social value.Sentiment analysis is a new technology of information mining, mainlyprocessing text emotion, attitude or opinion. In the process of sentiment analysis,a review will be classified as commendatory, derogatory or neutral withpersonal feelings, attitudes or opinions, also called text sentiment polarityclassification, is particularly important task in sentiment analysis.Commendatory comments authors tend to like comment object, derogatorycomments authors tend not to like comment object, neutral comments authorscan have a favorite tendency also have a tendency to not love, not have a cleartendency of emotion. This paper focuses on the sentiment analysis of Chineseparagraph-level. This paper firstly treats Chinese paragraph to be classified segmentation,clause, then classified subjective and objective sentence by the method ofmachine learning and extracted the subjective sentence with emotional color,removed the objective sentence does not contain emotional color. Aiming at thesubjective clues contains rich emotional information in subjective and objectiveclassification, especially the lack of study on associated words. This paperproposes a method which is characterized by subjective clues on subjective textrecognition, subjective clues ingredients as the subjective text recognition basis,use the Na ve Bayes classifier for Chinese paragraph on the classification ofsubjective and objective sentence, and extract the subjective sentence. Theexperimental results show that, using the proposed scheme can obtain a generalincrease in the accuracy of subjective text recognition by about8%, comparedto the method of Naive Bayes and subjective clues without associated words, byabout14%compared to the method of traditional Naive Bayes.Finally to extract the subjective emotional tendency analysis is based onthe analysis of emotional words, according to each subjective sentence inparagraph has different contributions to paragraph, this paper proposes a methodof text emotional propensity analysis based on the sentence weight, based on theemotional tendencies of all single sentences and all complex sentences havebeen determined, to synthesize emotional tendencies of all single sentences andall complex sentences by a scientific synthesis algorithm, then get the sentimentof the whole paragraph. The experiment shows that the method works well for paragraph-level of text sentiment classification.