EdinSaar@WMT21: North-Germanic Low-Resource Multilingual NMT

TL;DR

This paper presents multilingual translation models for North Germanic languages using various techniques, achieving superior performance in low-resource translation tasks at WMT2021.

Contribution

It introduces effective combination of multilingual pre-training, back-translation, fine-tuning, and ensembling for low-resource NMT.

Findings

Models outperform other submissions in most directions.

Techniques improve translation quality for Icelandic, Norwegian-Bokmal, and Swedish.

Demonstrates effectiveness of combined methods in low-resource settings.

Abstract

We describe the EdinSaar submission to the shared task of Multilingual Low-Resource Translation for North Germanic Languages at the Sixth Conference on Machine Translation (WMT2021). We submit multilingual translation models for translations to/from Icelandic (is), Norwegian-Bokmal (nb), and Swedish (sv). We employ various experimental approaches, including multilingual pre-training, back-translation, fine-tuning, and ensembling. In most translation directions, our models outperform other submitted systems.