PMTL: Personality-Aware Multitask Learning Model for Emotion Recognition in Conversation

In the field of affective computing, Emotion Recognition in Conversations (ERC) has gained increasing attention. The relationship between emotions and personality traits during conversations is a complex challenge. However, most conversational context models in ERC do not consider the speaker's background. Few works focus on personality traits and their impact on emotion recognition. To address this, we introduce PMTL, a multitask learning model that synergizes emotion recognition with sentiment analysis. PMTL incorporates personality traits into the emotional transition modeling process, leveraging the VAD emotional space to map contextual interdependencies and speaker-specific nuances. Our model outperforms existing models on benchmark datasets. Through an ablation study, we validate the efficacy of each component within PMTL, thereby demonstrating the significance of incorporating personality traits into emotion recognition tasks.