LIRMM-Advanse at SemEval-2019 Task 3: Attentive Conversation Modeling for Emotion Detection and Classification

Proceedings of the 13th International Workshop on Semantic Evaluation (SemEval-2019), pages 251–255 Minneapolis, Minnesota, USA, June 6–7, 2019. ©2019 Association for Computational Linguistics 251 LIRMM-Advanse at SemEval-2019 Task 3: Attentive Conversation Modeling for Emotion Detection and Classification Waleed Ragheb1,2 , Jérôme Azé 1,2 , Sandra Bringay1,3 and Maximilien Servajean1,3 1LIRMM UMR 5506, CNRS, University of Montpellier, Montpellier, France 2IUT de Béziers, University of Montpellier, Béziers, France 3AMIS, Paul Valery University - Montpellier 3 , Montpellier, France {First.Last}@lirmm.fr Abstract This paper addresses the problem of model- ing textual conversations and detecting emo- tions. Our proposed model makes use of 1) deep transfer learning rather than the clas- sical shallow methods of word embedding; 2) self-attention mechanisms to focus on the most important parts of the texts and 3) turn- based conversational modeling for classifying the emotions. Our model was evaluated on the data provided by the SemEval-2019 shared task on contextual emotion detection in text. The model shows very competitive results. 1 Introduction Emotional intelligence has played a signifi- cant role in many application in recent years (Krakovsky, 2018). It is one of the essential abilities to move from narrow to general human- like intelligence. Being able to recognize expres- sions of human emotion such as interest, distress, and pleasure in communication is vital for help- ing machines choose more helpful and less aggra- vating behavior. Human emotions are a mental state that can be sensed and hence recognized in many sources such as visual features in images or videos (Boubenna and Lee, 2018), as textual se- mantics and sentiments in texts (Calefato et al., 2017) or even patterns in EEG brain signals (Jenke et al., 2014). With the increasing number of mes- saging platforms and with the growing demand of customer chat bot applications, detecting the emotional state in conversations becomes highly important for more personalized and human-like conversations (Zhou et al., 2018). This paper addresses the problem of modeling a conversation that comes with multiple turns for detecting and classifying emotions. The proposed model makes use of transfer learning through the universal language modeling that is composed of consecutive layers of Bi-directional Long Term Short Term Memory (Bi-LSTM) units. These lay- ers are learned first in sequence-to-sequence fash- ion on a general text and then fine-tuned to a spe- cific target task. The model also makes use of an attention mechanism in order to focus on the most important parts of each text turn. Finally, the pro- posed classifier models the changing of the emo- tional state of a specific user across turns. The rest of the paper is organized as follows. In Section 2, the related work is introduced. Then, we present a quick overview of the task and the datasets in Section 3. Section 4 describes the pro- posed model architecture, some variants and hy- perparameters settings. The experiments and re- sults are presented in Section 5. Section 6 con- cludes the study. 2 Related Work Transfer learning or domain adaptation has been widely used in machine …

Content truncated to 500 words