Zero and Few-Shot Localization of Task-Oriented Dialogue Agents with a Distilled Representation

TL;DR

This paper introduces a novel end-to-end approach for zero and few-shot localization of task-oriented dialogue agents across languages, significantly reducing the need for target language training data while maintaining high accuracy.

Contribution

It presents methods to build high-quality dialogue agents in new languages using improved data representation, entity-aware translation, and filtering, outperforming prior cross-lingual approaches.

Findings

Zero-shot Task Success Rate: 46.7% in Chinese-English transfer.

Few-shot improvement: 15.2% over state-of-the-art.

Approach narrows gap to full-shot performance within 5%.

Abstract

Task-oriented Dialogue (ToD) agents are mostly limited to a few widely-spoken languages, mainly due to the high cost of acquiring training data for each language. Existing low-cost approaches that rely on cross-lingual embeddings or naive machine translation sacrifice a lot of accuracy for data efficiency, and largely fail in creating a usable dialogue agent. We propose automatic methods that use ToD training data in a source language to build a high-quality functioning dialogue agent in another target language that has no training data (i.e. zero-shot) or a small training set (i.e. few-shot). Unlike most prior work in cross-lingual ToD that only focuses on Dialogue State Tracking (DST), we build an end-to-end agent. We show that our approach closes the accuracy gap between few-shot and existing full-shot methods for ToD agents. We achieve this by (1) improving the dialogue data representation, (2) improving entity-aware machine translation, and (3) automatic filtering of noisy translations. We evaluate our approach on the recent bilingual dialogue dataset BiToD. In Chinese to English transfer, in the zero-shot setting, our method achieves 46.7% and 22.0% in Task Success Rate (TSR) and Dialogue Success Rate (DSR) respectively. In the few-shot setting where 10% of the data in the target language is used, we improve the state-of-the-art by 15.2% and 14.0%, coming within 5% of full-shot training.