Few-shot Natural Language Generation for Task-Oriented Dialog

TL;DR

This paper introduces FewShotWoz, a new benchmark for few-shot natural language generation in task-oriented dialog, and proposes SC-GPT, a model that effectively adapts to new domains with limited data, outperforming existing methods.

Contribution

It presents the first few-shot NLG benchmark for task-oriented dialog and develops SC-GPT, a controllable generation model that adapts with minimal domain-specific data.

Findings

SC-GPT outperforms existing methods on FewShotWoz and Multi-Domain-WOZ datasets.

SC-GPT achieves higher automatic and human evaluation scores.

Few-shot learning is effective for domain adaptation in NLG.

Abstract

As a crucial component in task-oriented dialog systems, the Natural Language Generation (NLG) module converts a dialog act represented in a semantic form into a response in natural language. The success of traditional template-based or statistical models typically relies on heavily annotated data, which is infeasible for new domains. Therefore, it is pivotal for an NLG system to generalize well with limited labelled data in real applications. To this end, we present FewShotWoz, the first NLG benchmark to simulate the few-shot learning setting in task-oriented dialog systems. Further, we develop the SC-GPT model. It is pre-trained on a large set of annotated NLG corpus to acquire the controllable generation ability, and fine-tuned with only a few domain-specific labels to adapt to new domains. Experiments on FewShotWoz and the large Multi-Domain-WOZ datasets show that the proposed SC-GPT significantly outperforms existing methods, measured by various automatic metrics and human evaluations.