CUNI Systems for the Unsupervised and Very Low Resource Translation Task in WMT20
This addresses machine translation for low-resource language pairs, showing incremental improvements using synthetic data and transfer learning.
The paper tackled unsupervised and very low-resource machine translation between German and Upper Sorbian, achieving 25.5 and 23.7 BLEU in unsupervised scenarios and 57.4 and 56.1 BLEU with low-resource systems, improving by 10 BLEU points over a baseline.
This paper presents a description of CUNI systems submitted to the WMT20 task on unsupervised and very low-resource supervised machine translation between German and Upper Sorbian. We experimented with training on synthetic data and pre-training on a related language pair. In the fully unsupervised scenario, we achieved 25.5 and 23.7 BLEU translating from and into Upper Sorbian, respectively. Our low-resource systems relied on transfer learning from German-Czech parallel data and achieved 57.4 BLEU and 56.1 BLEU, which is an improvement of 10 BLEU points over the baseline trained only on the available small German-Upper Sorbian parallel corpus.