Affective Decoding for Empathetic Response Generation

TL;DR

This paper introduces Affective Decoding, a technique that incorporates emotion signals during response generation to improve empathy in dialogue systems, validated through human evaluations showing increased perceived empathy.

Contribution

It proposes a novel Affective Decoding method with an auxiliary dual emotion encoder for more empathetic responses in dialogue systems.

Findings

Models are perceived as more empathetic by humans.

Affective Decoding effectively incorporates emotion signals during decoding.

Auxiliary dual emotion encoder enhances empathetic response quality.

Abstract

Understanding speaker's feelings and producing appropriate responses with emotion connection is a key communicative skill for empathetic dialogue systems. In this paper, we propose a simple technique called Affective Decoding for empathetic response generation. Our method can effectively incorporate emotion signals during each decoding step, and can additionally be augmented with an auxiliary dual emotion encoder, which learns separate embeddings for the speaker and listener given the emotion base of the dialogue. Extensive empirical studies show that our models are perceived to be more empathetic by human evaluations, in comparison to several strong mainstream methods for empathetic responding.