Research Of Microblogging Comments Emotion Classification Based On CRFs

In the information society, information transfer in different ways through weibo thisconvenient way of information exchange, information transmission has been deep into everycorner of our life. On the microblogging platform has tens of thousands of users, and oftenpublished on twitter for something or a hot topic of discussion and personal emotional point ofview. Preserved on the microblogging platform, therefore, a large number of corpus analysis, canbe found that most of the common crowd, emotion and value orientation, for providing the basisof the analysis problems concerned with issues related to policy makers.This article first to the analysis of the existing corpus emotion has carried on the inductionand summary of related research. Then, comparing the several common emotion classificationmodel, including the method based on similarity, a bayesian classifier and support vector machine(SVM), etc. Through the analysis of the advantages and disadvantages of each model, in the end,current widely accepted an emotion classification methods, conditions and field (CRFs);Secondly, using the words on the level of granularity characteristic of Chinese sentences in thetext, labeling, using conditions and field model of experimental training corpus, form a trainingmodel, using the trained model to review the information on the judgment of emotiontendentiousness. Finally, the paper puts forward a kind of emotional intensity classificationmechanism, makes the sentiment analysis is not just limited to the positive, neutral, and reversethree conditions, the experimental results to quantify the original three aspects, thus throughquantitative results, after ranking on the strength of the emotion.In this paper, by using the CRFs reached after the analysis of the corpus, CRFs for emotionalstatement has better classification effect, and apply the results basically of emotional strengthgrading mechanism proposed by the authors is verified the feasibility of, through quantitativeresults can provide data support for decision makers. But are still in the study of room forimprovement, such as corpus is still not very complete, will further improve in the future.