For the mass texts in the micro-blog, uses the dictionary of opinion words and the method of support vector machine (SVM) to recognize the opinion sentence in Chinese micro-blog. Constructs the dictionary of opinion words, counts five features of Chinese micro-blog opinion sentences, selects four features to recognize opinion sentences, as well as compares the SVM-based algorithm and the algorithm of opinion words dictionary. The contrast results show that the SVM-based method is best in identifying the micro-blog opinion sentence, and the accuracy is 68.75 %, the recall rate is 48.71 % and the F-measure is 57.02 %.