Resonating response makes people feel better: An empathetic protocol in dialogue system

Abstract

Currently, the emotional research of dialogue systems is a hot topic. However, several works mainly focused on acquiring state-of-the-art performance in a dialogue system and paid less attention to the inner emotions' response and lacked interpretability of emotional response mechanism within a dialogue system. Hence, this work proposed an empathic protocol to address this issue via introducing an innovative element (Mirror neuron) from connectionism and neuroscience to gradually design an AMNN (Artificial mirror neuron network) in the dialogue system for clear interpretability firstly. Subsequently, this paper described an empathic protocol to produce and analyze responses between a user and an agent via the self-defined neural network that served as the Central Nervous System of a dialogue agent. By employing this protocol in a traffic-service application, users felt that their emotions were resonated with and understood and communicated with the dialogue agent proactively.