@inproceedings{edman-etal-2021-unsupervised,
title = "Unsupervised Translation of {G}erman{--}{L}ower {S}orbian: Exploring Training and Novel Transfer Methods on a Low-Resource Language",
author = {Edman, Lukas and
{\"U}st{\"u}n, Ahmet and
Toral, Antonio and
van Noord, Gertjan},
booktitle = "Proceedings of the Sixth Conference on Machine Translation",
month = nov,
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.wmt-1.104",
pages = "982--988",
abstract = "This paper describes the methods behind the systems submitted by the University of Groningen for the WMT 2021 Unsupervised Machine Translation task for German{--}Lower Sorbian (DE{--}DSB): a high-resource language to a low-resource one. Our system uses a transformer encoder-decoder architecture in which we make three changes to the standard training procedure. First, our training focuses on two languages at a time, contrasting with a wealth of research on multilingual systems. Second, we introduce a novel method for initializing the vocabulary of an unseen language, achieving improvements of 3.2 BLEU for DE-{\textgreater}DSB and 4.0 BLEU for DSB-{\textgreater}DE.Lastly, we experiment with the order in which offline and online back-translation are used to train an unsupervised system, finding that using online back-translation first works better for DE-{\textgreater}DSB by 2.76 BLEU. Our submissions ranked first (tied with another team) for DSB-{\textgreater}DE and third for DE-{\textgreater}DSB.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="edman-etal-2021-unsupervised">
<titleInfo>
<title>Unsupervised Translation of German–Lower Sorbian: Exploring Training and Novel Transfer Methods on a Low-Resource Language</title>
</titleInfo>
<name type="personal">
<namePart type="given">Lukas</namePart>
<namePart type="family">Edman</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ahmet</namePart>
<namePart type="family">Üstün</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Antonio</namePart>
<namePart type="family">Toral</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gertjan</namePart>
<namePart type="family">van Noord</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2021-nov</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Sixth Conference on Machine Translation</title>
</titleInfo>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Online</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>This paper describes the methods behind the systems submitted by the University of Groningen for the WMT 2021 Unsupervised Machine Translation task for German–Lower Sorbian (DE–DSB): a high-resource language to a low-resource one. Our system uses a transformer encoder-decoder architecture in which we make three changes to the standard training procedure. First, our training focuses on two languages at a time, contrasting with a wealth of research on multilingual systems. Second, we introduce a novel method for initializing the vocabulary of an unseen language, achieving improvements of 3.2 BLEU for DE-\textgreaterDSB and 4.0 BLEU for DSB-\textgreaterDE.Lastly, we experiment with the order in which offline and online back-translation are used to train an unsupervised system, finding that using online back-translation first works better for DE-\textgreaterDSB by 2.76 BLEU. Our submissions ranked first (tied with another team) for DSB-\textgreaterDE and third for DE-\textgreaterDSB.</abstract>
<identifier type="citekey">edman-etal-2021-unsupervised</identifier>
<location>
<url>https://aclanthology.org/2021.wmt-1.104</url>
</location>
<part>
<date>2021-nov</date>
<extent unit="page">
<start>982</start>
<end>988</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Unsupervised Translation of German–Lower Sorbian: Exploring Training and Novel Transfer Methods on a Low-Resource Language
%A Edman, Lukas
%A Üstün, Ahmet
%A Toral, Antonio
%A van Noord, Gertjan
%S Proceedings of the Sixth Conference on Machine Translation
%D 2021
%8 nov
%I Association for Computational Linguistics
%C Online
%F edman-etal-2021-unsupervised
%X This paper describes the methods behind the systems submitted by the University of Groningen for the WMT 2021 Unsupervised Machine Translation task for German–Lower Sorbian (DE–DSB): a high-resource language to a low-resource one. Our system uses a transformer encoder-decoder architecture in which we make three changes to the standard training procedure. First, our training focuses on two languages at a time, contrasting with a wealth of research on multilingual systems. Second, we introduce a novel method for initializing the vocabulary of an unseen language, achieving improvements of 3.2 BLEU for DE-\textgreaterDSB and 4.0 BLEU for DSB-\textgreaterDE.Lastly, we experiment with the order in which offline and online back-translation are used to train an unsupervised system, finding that using online back-translation first works better for DE-\textgreaterDSB by 2.76 BLEU. Our submissions ranked first (tied with another team) for DSB-\textgreaterDE and third for DE-\textgreaterDSB.
%U https://aclanthology.org/2021.wmt-1.104
%P 982-988
Markdown (Informal)
[Unsupervised Translation of German–Lower Sorbian: Exploring Training and Novel Transfer Methods on a Low-Resource Language](https://aclanthology.org/2021.wmt-1.104) (Edman et al., WMT 2021)
ACL