Sentiment classification on weibo has recently attracted wide attention in research community. Most previous works are focused on weibo comments regarding movies or products. Our study, in contrast, is aimed at gusty incidents on weibo. Comments of the incidents are considered either positive or negative representing attitudes of users towards these incidents. Classifying users’ attitudes helps identifying the general opinion of the public. In this paper, we propose an innovative convolutional neural networks based method, termed as CNN-SVM, to classify the incident comments. In addition, according to users’ repost actions, we propose a new data structure, repost tree, for dealing with ambiguity in the comments. Extensive experiments demonstrate that the CNN-SVMmethod effectively improves the accuracy of incidents sentiment classification. The new data structure shows to be effective on steering the classification results towards real world sentiment tendency.
[1]
Yee Whye Teh,et al.
A Fast Learning Algorithm for Deep Belief Nets
,
2006,
Neural Computation.
[2]
Yoon Kim,et al.
Convolutional Neural Networks for Sentence Classification
,
2014,
EMNLP.
[3]
Tong Zhang,et al.
Effective Use of Word Order for Text Categorization with Convolutional Neural Networks
,
2014,
NAACL.
[4]
Bo Pang,et al.
Thumbs up? Sentiment Classification using Machine Learning Techniques
,
2002,
EMNLP.
[5]
Qingcai Chen,et al.
Fuzzy deep belief networks for semi-supervised sentiment classification
,
2014,
Neurocomputing.
[6]
Yoshua Bengio,et al.
Gradient-based learning applied to document recognition
,
1998,
Proc. IEEE.
[7]
Philip Resnik,et al.
Political Ideology Detection Using Recursive Neural Networks
,
2014,
ACL.