An Exploratory Analysis of Multilingual Word-Level Quality Estimation with Cross-Lingual Transformers

TL;DR

This paper investigates multilingual word-level Quality Estimation using cross-lingual transformers, demonstrating that such models can perform comparably to language-specific models and generalize well across different language pairs, including zero-shot and few-shot scenarios.

Contribution

It introduces multilingual QE models based on pre-trained transformers that eliminate the need for language-specific training data, enabling effective cross-lingual predictions.

Findings

Multilingual QE models perform on par with language-specific models.

Zero-shot and few-shot QE are feasible with the proposed models.

Transformer-based models generalize well across languages.

Abstract

Most studies on word-level Quality Estimation (QE) of machine translation focus on language-specific models. The obvious disadvantages of these approaches are the need for labelled data for each language pair and the high cost required to maintain several language-specific models. To overcome these problems, we explore different approaches to multilingual, word-level QE. We show that these QE models perform on par with the current language-specific models. In the cases of zero-shot and few-shot QE, we demonstrate that it is possible to accurately predict word-level quality for any given new language pair from models trained on other language pairs. Our findings suggest that the word-level QE models based on powerful pre-trained transformers that we propose in this paper generalise well across languages, making them more useful in real-world scenarios.