Interactive Emotional Learning in Emotional Intelligent Dialogue Systems
Qiang Li, Feng Zhao, Hong Ouyang, Hui Xiang, Jianning Zheng, Linlin Zhao, Hongbing Zhang, Ting Guo · 2024
With the continuous advancement of artificial intelligence in the field of conversational systems, users have raised higher demands for the emotional expression capabilities of dialogue agents. To address this issue, we proposed an Interactional Emotional Learning model (TIEL) tailored for task-oriented dialogue, aimed at enhancing the emotional prediction accuracy of dialogue agents and addressing the issue of missing target text. The TIEL model achieved a deep understanding of multi-turn dialogue contexts and accurate prediction of emotional categories through three modules: knowledge-enhanced dialogue context encoding, latent emotional response learning, and interactive emotional prediction. Experimental results on the DailyDialog dataset demonstrated a significant performance improvement of the TIEL model in emotion prediction tasks.