IASBS at SemEval-2024 Task 10: Delving into Emotion Discovery and Reasoning in Code-Mixed Conversations
Mehrzad Tareh, Aydin Mohandesi, Ebrahim Ansari · 2024
In this study, we introduce the SemEval 2024 Task 10, entitled "Emotion Discovery and Reasoning its Flip in Conversation (EDiReF)".Our research presents a comprehensive framework to analyze emotional dynamics within Hindi-English mixed-language and English conversations.We extend beyond traditional emotion identification to uncover the triggers behind shifts in emotional states using advanced Natural Language Processing (NLP) techniques.Employing a systematic methodology that encompasses data preprocessing, feature engineering, and the deployment of language models such as GPT-4 and DistilBERT, we unravel the complex interplay of emotions in communication.Our approach yields significant insights, enhancing applications from social media analytics to mental health, thus marking a notable advancement in the integration of emotional intelligence into AI.Noteworthy is our system's achievement of third place on the leaderboard, demonstrating robust performance with a weighted F1-Score of 0.70.This study not only contributes to the field of emotional AI but also paves the way for future research on the nuanced understanding of emotion in mixed-language communications.