Effective Data Augmentation Approaches to End-to-End Task-Oriented Dialogue

TL;DR

This paper introduces four automatic data augmentation methods at word and sentence levels for end-to-end task-oriented dialogue systems, significantly improving performance and robustness without costly manual annotation.

Contribution

It presents novel automatic data augmentation approaches and demonstrates their effectiveness in enhancing dialogue system performance on benchmark datasets.

Findings

All four augmentation methods improve Success F1 scores.

Ensemble of methods achieves state-of-the-art results.

Methods increase diversity and robustness of user utterances.

Abstract

The training of task-oriented dialogue systems is often confronted with the lack of annotated data. In contrast to previous work which augments training data through expensive crowd-sourcing efforts, we propose four different automatic approaches to data augmentation at both the word and sentence level for end-to-end task-oriented dialogue and conduct an empirical study on their impact. Experimental results on the CamRest676 and KVRET datasets demonstrate that each of the four data augmentation approaches is able to obtain a significant improvement over a strong baseline in terms of Success F1 score and that the ensemble of the four approaches achieves the state-of-the-art results in the two datasets. In-depth analyses further confirm that our methods adequately increase the diversity of user utterances, which enables the end-to-end model to learn features robustly.