CNN-SVM with Embedded Recurrent Structure for Social Emotion Prediction
Rida Dong, Oinke Peng, Xintong Li, Xinyu Guan · 2018
The rapid development of the Internet has generated a large amount of online user-generated information. Automatic sentiment analysis of the user-generated information has great research and application prospects. Traditional sentiment analysis task mainly focuses on authors' emotions. Instead, our research aims at the emotions of readers invoked by news articles, which are called social emotions. In this paper, we propose a novel method, CNN-SVM with Embedded Recurrent Structure, for social emotion prediction. Specifically, we replace the fixed window convolutional layer in CNN with a bidirectional recurrent structure, that is, our model is a cascade of the bidirectional recurrent structure and a max-pooling layer. Then, the values of max-pooling layer are used as extracted features to predict social emotions with a SVM classifier. Experimental results show that our method outperforms the state-of-the-art methods in social emotion prediction by a significant margin.