Multi-Domain Dialogue Acts and Response Co-Generation

TL;DR

This paper introduces a neural co-generation model for task-oriented dialogue systems that simultaneously generates dialogue acts and responses, preserving semantic structures and dynamically attending to relevant acts, leading to improved performance.

Contribution

The paper presents a novel joint generation model that captures semantic structures of multi-domain dialogue acts and enhances response quality through dynamic attention, outperforming existing pipeline methods.

Findings

Significant improvement over state-of-the-art models in automatic evaluations.

Enhanced response relevance and informativeness demonstrated in human assessments.

Effective joint training with uncertainty loss improves task balance.

Abstract

Generating fluent and informative responses is of critical importance for task-oriented dialogue systems. Existing pipeline approaches generally predict multiple dialogue acts first and use them to assist response generation. There are at least two shortcomings with such approaches. First, the inherent structures of multi-domain dialogue acts are neglected. Second, the semantic associations between acts and responses are not taken into account for response generation. To address these issues, we propose a neural co-generation model that generates dialogue acts and responses concurrently. Unlike those pipeline approaches, our act generation module preserves the semantic structures of multi-domain dialogue acts and our response generation module dynamically attends to different acts as needed. We train the two modules jointly using an uncertainty loss to adjust their task weights adaptively. Extensive experiments are conducted on the large-scale MultiWOZ dataset and the results show that our model achieves very favorable improvement over several state-of-the-art models in both automatic and human evaluations.