Relevant Emotion Ranking from Text Constrained with Emotion Relationships

Proceedings of NAACL-HLT 2018, pages 561–571 New Orleans, Louisiana, June 1 - 6, 2018. c⃝2018 Association for Computational Linguistics Relevant Emotion Ranking from Text Constrained with Emotion Relationships Deyu Zhou† Yang Yang† Yulan He§ † School of Computer Science and Engineering, Key Laboratory of Computer Network and Information Integration, Ministry of Education, Southeast University, China § School of Engineering and Applied Science, Aston University, UK {d.zhou,yyang}@seu.edu.cn, y.he@cantab.net Abstract Text might contain or invoke multiple emo- tions with varying intensities. As such, emo- tion detection, to predict multiple emotions as- sociated with a given text, can be cast into a multi-label classification problem. We would like to go one step further so that a ranked list of relevant emotions are generated where top ranked emotions are more intensely associat- ed with text compared to lower ranked emo- tions, whereas the rankings of irrelevant emo- tions are not important. A novel framework of relevant emotion ranking is proposed to tackle the problem. In the framework, the objective loss function is designed elaborately so that both emotion prediction and rankings of on- ly relevant emotions can be achieved. More- over, we observe that some emotions co-occur more often while other emotions rarely co- exist. Such information is incorporated into the framework as constraints to improve the accuracy of emotion detection. Experimental results on two real-world corpora show that the proposed framework can effectively deal with emotion detection and performs remark- ably better than the state-of-the-art emotion detection approaches and multi-label learning methods. 1 Introduction With the growing prosperity of Web 2.0, people tend to share their feelings, attitudes and opin- ions through the social platforms such as online news sites, blogs. Detecting emotions from tex- t can enhance the understanding of users’ emo- tional states, which is useful in many downstream applications, such as human-computer interaction and personalized recommendation. Therefore, it is crucial to analyze and predict emotions from text accurately (Picard and Picard, 1997). Research on emotion detection can be rough- ly categorized into two types: lexicon-based and learning-based approaches. Lexicon-based ap- proaches usually rely on emotion lexicons (Lei et al., 2014; Rao et al., 2012). They cannot deal with text when words can’t be found in emo- tion lexicons. Learning-based approaches can be furthered classified into unsupervised and super- vised learning methods. Unsupervised approach- es do not require annotated data for training. For example, by adding an emotion layer into tradi- tional topic models, emotion-topic models were constructed to detect users’ emotions (Bao et al., 2012, 2009). Supervised learning approaches con- sider each emotion category as a class label and emotion detection is cast as a classification prob- lem. If only choosing the strongest emotion as the emotion label for a given text, emotion detection is essentially a single-label classification prob- lem (Lin et al., 2008; Quan et al., 2015). To predic- t multiple emotions simultaneously, emotion de- tection can be solved in the multi-label classifica- tion framework (Bhowmick, 2009). Moreover, to predict both multiple emotions and their intensi- ties, some …

Content truncated to 500 words