Event Based Emotion Classification for News Articles

Reading of news articles can trigger emotional reactions from its readers. But comparing to other genre of text, news articles that are mainly used to report events, lack emotion linked words and other features for emotion classification. In this paper, we propose an event anchor based method for emotion classification for news articles. Firstly, we build an emotion linked news corpus through crowdsourcing. Then we propose a CRF based event anchor extraction method to identify event related anchor words that can potentially trigger emotions. These anchor words are then used as features to train a classifier for emotion classification. Experiment shows that our proposed anchor word based method achieves comparable performance to bag-ofword based method and it also performs better than emotion lexicon features. Combining anchor words with bag-of-words can increase the performance by 7.0% under weighted Fscore. Evaluation on the SemEval 2007 news headlines task shows that our method outperforms most of other methods.