English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

This paper describes the methods behind the systems submitted by theUniversity of Groningen for the WMT 2021 Unsupervised Machine Translation taskfor German--Lower Sorbian (DE--DSB): a high-resource language to a low-resourceone. Our system uses a transformer encoder-decoder architecture in which wemake three changes to the standard training procedure. First, our trainingfocuses on two languages at a time, contrasting with a wealth of research onmultilingual systems. Second, we introduce a novel method for initializing thevocabulary of an unseen language, achieving improvements of 3.2 BLEU forDE$\rightarrow$DSB and 4.0 BLEU for DSB$\rightarrow$DE. Lastly, we experimentwith the order in which offline and online back-translation are used to trainan unsupervised system, finding that using online back-translation first worksbetter for DE$\rightarrow$DSB by 2.76 BLEU. Our submissions ranked first (tiedwith another team) for DSB$\rightarrow$DE and third for DE$\rightarrow$DSB.

Unsupervised Translation of German--Lower Sorbian: Exploring Training and Novel Transfer Methods on a Low-Resource Language