Multi-Task Learning for Situated Multi-Domain End-to-End Dialogue Systems

TL;DR

This paper introduces a multi-task learning approach using GPT-2 for multi-domain, multi-modal dialogue systems, achieving superior performance over specialized models and demonstrating the effectiveness of various training strategies.

Contribution

It presents a novel multi-task training framework for GPT-2 that handles complex, multi-domain dialogue tasks with improved results and comprehensive ablation studies.

Findings

Single GPT-2 model outperforms domain-specific models

Multi-task learning enhances dialogue system performance

Proposed strategies further improve GPT-2 based dialogue systems

Abstract

Task-oriented dialogue systems have been a promising area in the NLP field. Previous work showed the effectiveness of using a single GPT-2 based model to predict belief states and responses via causal language modeling. In this paper, we leverage multi-task learning techniques to train a GPT-2 based model on a more challenging dataset with multiple domains, multiple modalities, and more diversity in output formats. Using only a single model, our method achieves better performance on all sub-tasks, across domains, compared to task and domain-specific models. Furthermore, we evaluated several proposed strategies for GPT-2 based dialogue systems with comprehensive ablation studies, showing that all techniques can further improve the performance.