The Detection Of Nonsense And Abnormity In Comments

In recent years,a large amount of comments which contain opinion and emotional tendency have arisen because of the appearance of Social Media. It has become more and more valuable for organizaions to know the public feelings by mining rebiew. However, too much meaningless comments and abnormal expression of emotion make this task difficult.Based on this background, this paper studies the detection of nonsense and abnormity in comments. This research aims to detect the comments which don’t contain reviewers’ subjective opinion, and to detect the comments which contain sarcasm as an abnormal emotional expression. The main work of this paper is as the following:1. The research of this part is to detect nonsense in comments. Based on the defination of nonsense proposed by this paper, we design some useful features and build a classifier to implement the detection of nonsense in comments.2. The research of this part is to detect sarcasm from twitter comments based on features. To extract context features,this paper gives the defination of context based on frequency which can make the number of context comments change dynamcally. Then this paper improved the existed way and designs a new way to extract context features.3. The research of this part is to detect sarcasm from forum comments based on features. To extract context features, this paper proposes the context features inside comments and the topic features. To extract text features, this paper transferred some useful features from twitter data to detect sarcasm emotion.