Emotion Translator in Conversational AI Based on Big Five Personality Profiles
Fuad Zaki Nurrahman, Kosuke Takano, Hestiasari Rante · 2025
Traditional chatbots often lack emotional awareness, resulting in interactions that feel generic, disconnected, and lacking human-like depth. This limitation reduces their effectiveness in emotionally sensitive or interactive storytelling applications. To address this issue, this study proposes an enhanced Emotional AI framework that integrates personality traits-specifically the Big Five Personality Traits- in to the emotional response mechanism. The system employs a fine-tuned Large Language Model (LLM) alongside a static personality-emotion matrix to modulate emotional updates in a way that reflects the AI character's individual personality. For instance, a character high in Openness maintains stable, optimistic emotional patterns, while one high in Neuroticism exhibits more volatile emotional responses. Implemented within the Ren'Py Visual Novel Engine, the method enables real-time emotional expression through dynamic character sprite transitions and dialogue generation. Experimental results show that the approach improves emotional stability by 60%, alignment with personality traits by 35%, and dialogue coherence by 75% compared to a baseline model. These findings demonstrate the system's effectiveness in producing emotionally consistent, personality-aligned AI companions, advancing the development of emotionally intelligent storytelling agents.