Research of the Social Comment Data Emotional Tendency Based on Object Select SVM-Text Sentiment Model
Wenxin Hu, Liang He · International Journal of Information and Computer Science · 2016
In the study of hits of the drama, the useful of the social comment data is more and more important. But the most research concentrated on the volume or the kinds of the social data. The research of the emotional tendency indicated from the social date is insufficient. In this paper, we demonstrate that (1) because of the difference between drama social comment data and other comment, it’s necessary to build an emotional vocabulary for it; (2) confirming the object of the comment can improve the accuracy of result; (3) the emotion reverse is common in social comment data, it affects the real emotional tendency; (4) positive and negative words play different roles in the comment, they should be treated with different weight. In consideration of all the features extracted from Social comment data, we use the model of object Select SVM-Text Sentiment to get the accurate emotional tendency. Compared with the HowNet and NTUSD, the veracity is improved by 10% and 12% respectively; with confirming the comment object, the accuracy is improved nearly 7.1%; compared with the NBM and SVM,OSS-TS has the advantage of 14% and 15% respectively.