The past research of text emotion classification mainly face to the authors emotion.But this article classify textsemotion from the readers perspective.Using news articles as training sample sets

and using readersvote information behind news articles as prior knowledge of the sample category.we put forward a semi-supervised classification model according to the incomplete data sets

the classification method using naive bayesian method and in couple with EM algorithm.Experiments show that our method is not only simple and effective

and has higher classification performance.