Many Hands Make Light Work: Task-Oriented Dialogue System with Module-Based Mixture-of-Experts

TL;DR

This paper introduces SMETOD, a module-based mixture-of-experts approach for task-oriented dialogue systems that improves performance and efficiency by specializing subcomponents, outperforming existing models on key benchmarks.

Contribution

The paper proposes a novel Soft Mixture-of-Experts framework for dialogue systems, enhancing scalability, flexibility, and inference efficiency while achieving state-of-the-art results.

Findings

SMETOD outperforms baseline models on intent prediction, dialogue state tracking, and response generation.

SMETOD maintains high inference efficiency with reduced computational costs.

Experimental results demonstrate superior accuracy and problem-solving ability.

Abstract

Task-oriented dialogue systems are broadly used in virtual assistants and other automated services, providing interfaces between users and machines to facilitate specific tasks. Nowadays, task-oriented dialogue systems have greatly benefited from pre-trained language models (PLMs). However, their task-solving performance is constrained by the inherent capacities of PLMs, and scaling these models is expensive and complex as the model size becomes larger. To address these challenges, we propose Soft Mixture-of-Expert Task-Oriented Dialogue system (SMETOD) which leverages an ensemble of Mixture-of-Experts (MoEs) to excel at subproblems and generate specialized outputs for task-oriented dialogues. SMETOD also scales up a task-oriented dialogue system with simplicity and flexibility while maintaining inference efficiency. We extensively evaluate our model on three benchmark functionalities: intent prediction, dialogue state tracking, and dialogue response generation. Experimental results demonstrate that SMETOD achieves state-of-the-art performance on most evaluated metrics. Moreover, comparisons against existing strong baselines show that SMETOD has a great advantage in the cost of inference and correctness in problem-solving.