KGPChamps at SemEval-2019 Task 3: A deep learning approach to detect emotions in the dialog utterances.

Proceedings of the 13th International Workshop on Semantic Evaluation (SemEval-2019), pages 241–246 Minneapolis, Minnesota, USA, June 6–7, 2019. ©2019 Association for Computational Linguistics 241 KGPChamps at SemEval-2019 Task 3: A deep learning approach to detect emotions in the dialog utterances. 1Jasabanta Patro, 2Nitin Choudhary, 3Kalpit Chittora, and 4Animesh Mukherjee IIT Kharagpur, India {1jasabantapatro, 2nitch13jan, 3chittora.kalpit }@iitkgp.ac.in, 4animeshm@cse.iitkgp.ac.in Abstract This paper describes our approach to solve Se- meval task 3: EmoContext; where, given a tex- tual dialogue, i.e., a user utterance along with two turns of context, we have to classify the emotion associated with the utterance as one of the following emotion classes: Happy, Sad, Angry or Others. To solve this problem, we ex- periment with different deep learning models ranging from simple LSTM to relatively more complex attention with Bi-LSTM model. We also experiment with word embeddings such as ConceptNet along with word embeddings generated from bi-directional LSTM taking in- put characters. We fine tune different param- eters and hyper-parameters associated with each of our model and report the micro pre- cision, micro recall and micro F1-score for each model. We identify the Bi-LSTM model, along with the input word embedding taken as the concatenation of the embeddings gen- erated from the bidirectional character LSTM and ConceptNet embedding, as the best per- forming model with a highest micro-F1 score over the test set as 0.7261. 1 Introduction In recent years, with the increase in the popularity of social media platforms, a significant amount of unstructured social media content (posts, tweets, messages etc.) has become available to the re- search community. People use social media as a platform to share their opinions, emotions, thoughts etc. This information has a huge poten- tial to serve as a commercial catalyst to the busi- ness of companies and organizations, e.g., know- ing the opinion of people about a product or a ser- vice could help the company to do betterment of their product or service according to the desire of the online consumers. In similar lines, emotions from the peoples’ comments/opinion can help us to model the future popularity of the product or the service. Further, knowing public emotions about different events can help political parties to set their agenda for elections. Thus mining of opin- ions and emotions has a lot of practical relevance. Even prior to the social media era, emotion detec- tion had achieved significant attention of psychol- ogists and linguistics. An elaborate discussion of emotion as a research topic is presented in the next section. In this paper, we describe our system and the models, with which, we achieved signifi- cant performance improvement over the SemEval baseline for task 3. The task is described in (Chatterjee et al., 2019), where, given a textual dialogue, i.e., a user utterance along with two turns of context, we have to classify the emotion asso- ciated with the utterance into one of the following emotion classes: Happy, Sad, Angry or Others. To solve this problem, we experiment with different deep learning models ranging …

Content truncated to 500 words