Current approaches to empathetic response generation view the set of emotions
expressed in the input text as a flat structure, where all the emotions are
treated uniformly. We argue that empathetic responses often mimic the emotion
of the user to a varying degree, depending on its posi
Combining Trait and State emotions for Empathetic Response Model (CTSM) enhances emotional perception and empathetic expression in dialogue systems, outperforming state-of-the-art baselines.