Towards Multilingual Automatic Dialogue Evaluation

TL;DR

This paper addresses the challenge of developing multilingual dialogue evaluation metrics by leveraging multilingual pretrained models and data augmentation through machine translation, emphasizing the importance of translation quality control.

Contribution

It introduces a data augmentation method using machine translation and quality estimation to improve multilingual dialogue evaluation models.

Findings

Naive fine-tuning with translated data underperforms baseline.

Careful selection of translated data improves model performance.

Quality estimation metrics are crucial for effective data augmentation.

Abstract

The main limiting factor in the development of robust multilingual dialogue evaluation metrics is the lack of multilingual data and the limited availability of open sourced multilingual dialogue systems. In this work, we propose a workaround for this lack of data by leveraging a strong multilingual pretrained LLM and augmenting existing English dialogue data using Machine Translation. We empirically show that the naive approach of finetuning a pretrained multilingual encoder model with translated data is insufficient to outperform the strong baseline of finetuning a multilingual model with only source data. Instead, the best approach consists in the careful curation of translated data using MT Quality Estimation metrics, excluding low quality translations that hinder its performance.