IECCK: integrating emotion cause and common knowledge for empathetic dialogue generation

With the rapid development of dialogue systems, generating empathetic dialogue has become a research hotspot. However, existing methods often focus only on surface-level emotional information, neglecting the underlying causes of emotions and cognitive factors, leading to responses that lack genuine empathy. To address this, we propose an empathetic dialogue generation model that integrates emotion cause and common knowledge (IECCK). The model consists of three main modules. The Emotion Cause Extraction Module deeply explores the causes behind user emotions. The Common Knowledge Reasoning Module incorporates relevant common knowledge to consider cognitive factors. The Empathetic Dialogue Generation Module combines these elements to generate more empathetic responses. We also introduce a gated attention mechanism during decoding to enhance the focus on emotion cause words. Experimental results on the Empathetic Dialogues dataset show that our model achieves a perplexity of 35.16 and an emotion classification accuracy of 39.78 \(\%\) $\%$ , outperforming some baseline models.