Sentiment and Emotion help Sarcasm? A Multi-task Learning Framework for Multi-Modal Sarcasm, Sentiment and Emotion Analysis

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 4351–4360 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 4351 Sentiment and Emotion help Sarcasm? A Multi-task Learning Framework for Multi-Modal Sarcasm, Sentiment and Emotion Analysis Dushyant Singh Chauhan, Dhanush S R, Asif Ekbal and Pushpak Bhattacharyya Department of Computer Science & Engineering Indian Institute of Technology Patna Patna, Bihar, India-801106 {1821CS17, dhanush.cs16, asif, pb}@iitp.ac.in Abstract In this paper, we hypothesize that sarcasm is closely related to sentiment and emotion, and thereby propose a multi-task deep learning framework to solve all these three problems si- multaneously in a multi-modal conversational scenario. We, at first, manually annotate the recently released multi-modal MUStARD sarcasm dataset with sentiment and emotion classes, both implicit and explicit. For multi- tasking, we propose two attention mechanisms, viz. Inter-segment Inter-modal Attention (Ie- Attention) and Intra-segment Inter-modal At- tention (Ia-Attention). The main motivation of Ie-Attention is to learn the relationship be- tween the different segments of the sentence across the modalities. In contrast, Ia-Attention focuses within the same segment of the sen- tence across the modalities. Finally, repre- sentations from both the attentions are con- catenated and shared across the five classes (i.e., sarcasm, implicit sentiment, explicit sen- timent, implicit emotion, explicit emotion) for multi-tasking. Experimental results on the ex- tended version of the MUStARD dataset show the efficacy of our proposed approach for sar- casm detection over the existing state-of-the- art systems. The evaluation also shows that the proposed multi-task framework yields bet- ter performance for the primary task, i.e., sar- casm detection, with the help of two secondary tasks, emotion and sentiment analysis. 1 Introduction Sarcasm is an essential aspect of daily conversa- tion, and it adds more fun to the language. Oscar Wilde, an Irish poet-playwright, quotes, “Sarcasm is the lowest form of wit, but the highest form of intelligence”. Irrespective of its relation with intel- ligence, sarcasm is often challenging to understand. Sarcasm is often used to convey thinly veiled disapproval humorously. This can be easily de- picted through the following example, “This is so good, that I am gonna enjoy it in the balcony. I can enjoy my view, whilst I enjoy my dessert.” This ut- terance, at an outer glance, conveys that the speaker is extremely pleased with his dessert and wants to elevate the experience by enjoying it in the balcony. But, careful observation of the sentiment and emo- tion of the speaker helps us understand that the speaker is disgusted with the dessert and has a neg- ative sentiment during the utterance (c.f. Figure 1). This is where sentiment and emotion come into the picture. Sentiment, emotion and sarcasm are highly intertwined, and one helps in the understanding of the others better. "This is so good, that I am gonna enjoy it in the balcony. I can enjoy my view, whilst I enjoy my desert." Figure 1: Example to show that sentiment and emotion of the speaker can influence sarcasm detection Even though sentiment, emotion, …

Content truncated to 500 words