Teacher-Student Framework Enhanced Multi-domain Dialogue Generation

TL;DR

This paper introduces a teacher-student framework for multi-domain dialogue generation that eliminates the need for external state trackers, leveraging domain-specific teachers to train a universal student model directly on raw utterances.

Contribution

The novel framework trains a universal dialogue model without external state trackers by transferring knowledge from domain-specific teacher models to a student model using raw utterances.

Findings

Outperforms belief tracker-based systems in experiments

Effectively merges domain-specific knowledge into a universal model

Reduces error propagation in dialogue state tracking

Abstract

Dialogue systems dealing with multi-domain tasks are highly required. How to record the state remains a key problem in a task-oriented dialogue system. Normally we use human-defined features as dialogue states and apply a state tracker to extract these features. However, the performance of such a system is limited by the error propagation of a state tracker. In this paper, we propose a dialogue generation model that needs no external state trackers and still benefits from human-labeled semantic data. By using a teacher-student framework, several teacher models are firstly trained in their individual domains, learn dialogue policies from labeled states. And then the learned knowledge and experience are merged and transferred to a universal student model, which takes raw utterance as its input. Experiments show that the dialogue system trained under our framework outperforms the one uses a belief tracker.