Research On Sentiment Analysis Of Microblog Topics Based On Content Mining

Microblog as a new social networking media, with the timeliness of its release and ease of reading, has become the main way people obtain information and communication. Microblog contains multiple topics. Research on sentiment analysis of Microblog topics can assist the government to monitor public opinions, and it can help the enterprise to collect suggestions, maintain customer relationships, microblog marketing and crisis response and so on. So Sentiment analysis of Microblog topics is becoming a new hot research. Sentiment analysis of Microblog topics generally involves three tasks:the first is subjective microblog recognition; the second is sentiment orientation analysis; the third is the evaluation object extraction. But in fact, the study found that the emotional analysis of microblog users is useful, so this paper studies subjective microblog recognition, sentiment orientation analysis and emotional analysis.In subjective microblog recognition, the sentence, sentence structure and implicit were selected to represent the microblog. The method to recognition based on SVM model was proposed, and the cropus of NLP&CC2012 was used to test the model. the precision of the model is more than 0.87, the recall rate is moer than 79%. Selecting the same features, the SVM model is superior to CRFs model. The results of experiments shows remarkable performance of the proposed method.In sentiment orientation analysis, the emotion, sentence and the relationship between the sentences were proposed to represent the microblog. The SVM model was used to the task of sentiment orientation analysis. The proposed mehod was attended to the evalation of COAE2014. The results proves the effectiveness of the method.The precision is 96.1%, which is the first result in the evaluation, the recall is 44.2%, and F-measure is 60.1%,In emotional analysis, the mocriblog emotion are divided into five categories,which are "happiness", "anger", "sadness", "fear", "evil". Then the emotional ontology of Dalian University of Technology is improved, the SVM model is used to analyse the emotion. The cropus of NLP&CC2013 was used to testing the proposed model. The results proves the effectiveness of the method.