Abstract:Empathetic response generation endeavours to perceive the interlocutor’s emotional and cognitive states in the dialogue and express proper responses. Previous studies detect the interlocutor’s states by understanding the immediate context of the dialogue. However, these methods are at an elementary/intermediate level of empathetic understanding due to the neglect of the broader context (i.e., the situation) and its associations with the dialogue, leading to inaccurate comprehension of the interlocutor’s states. In this paper, we utilize the EMPATHETIC-DIALOGUES dataset consisting of 25k dialogues, and on this basis, we propose a Situation-Dialogue Association Model (SDAM). SDAM focuses on the broader context, i.e., the situation, and enhances the understanding of empathy from explicit and implicit associations. Regarding explicit associations, we propose a bidirectional filtering encoder. It selects relevant keywords between the situation and dialogue, learning their direct lexical relevance. For implicit associations, we use a knowledge-based hypergraph network grounded to learn convoluted connections between the situation and the dialogue. Moreover, we also introduce a simple fine-tuning approach that combines SDAM with large language models to further strengthen the empathetic understanding capability. Compared to the baseline, SDAM demonstrates superior empathetic ability. In terms of emotion accuracy, fluency, and response diversity (Distinct-1/Distinct-2), SDAM achieves improvements of 12.25 (a 30.47% increase), 0.3 (a 0.85% increase), and 0.86/1.23 (116.22% and 30.67% increases), respectively. Additionally, our variant model based on large language models exhibits better emotion recognition capability without compromising response quality, specifically achieving an improvement of 0.23 (a 0.37% increase) in emotion accuracy.

Fine-Grained Knowledge Enhancement for Empathetic Dialogue Generation.

Knowledge Bridging for Empathetic Dialogue Generation

Towards Empathetic Dialogue Generation over Multi-type Knowledge.

An Empathetic Dialogue Generation Model Based on External Knowledge Fusion

Empathetic Dialogue Generation via Sensitive Emotion Recognition and Sensible Knowledge Selection

Improving Empathetic Dialogue Generation by Dynamically Infusing Commonsense Knowledge.

TriKF: Triple-Perspective Knowledge Fusion Network for Empathetic Question Generation

Grafting Fine-tuning and Reinforcement Learning for Empathetic Emotion Elicitation in Dialog Generation

Improving Empathetic Response Generation by Emotion Recognition and Information Filtration

KnowDT: Empathetic dialogue generation with knowledge enhanced dependency tree

EmpDG: Multiresolution Interactive Empathetic Dialogue Generation

CAB: Empathetic Dialogue Generation with Cognition, Affection and Behavior

Improving Empathetic Dialogue Generation with Semantics Decoupling.

EmpCRL: Controllable Empathetic Response Generation via In-Context Commonsense Reasoning and Reinforcement Learning

Apeak-CG: Automatically predicting emotion based dynamic multi-form knowledge fusion conversation generation

A Knowledge-Enriched Model for Emotional Conversation Generation.

Empathetic Response Generation with Relation-aware Commonsense Knowledge.

Situation-aware Empathetic Response Generation

E-CORE: Emotion Correlation Enhanced Empathetic Dialogue Generation

Empathetic Dialog Generation with Fine-Grained Intents

Survey of Research on End-to-End Emotional Dialogue Generation