The University of Edinburgh’s English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task

Rachel Bawden; Alexandra Birch; Radina Dobreva; Arturo Oncevay Marcos; Antonio Valerio  Miceli Barone; Philip Williams

The University of Edinburgh’s English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task

Rachel Bawden, Alexandra Birch, Radina Dobreva, Arturo Oncevay Marcos, Antonio Valerio Miceli Barone, Philip Williams

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

We describe the University of Edinburgh's submissions to the WMT20 news translation shared task for the low resource language pair English-Tamil and the mid-resource language pair English-Inuktitut. We use the neural machine translation transformer architecture for all submissions and explore a variety of techniques to improve translation quality to compensate for the lack of parallel training data. For the very low-resource English-Tamil, this involves exploring pretraining, using both language model objectives and translation using an unrelated high-resource language pair (German-English), and iterative backtranslation. For English-Inuktitut, we explore the use of multilingual systems, which, despite not being part of the primary submission, would have achieved the best results on the test set.

Original language	English
Title of host publication	Proceedings of the 5th Conference on Machine Translation
Publisher	Association for Computational Linguistics (ACL)
Pages	92-99
Number of pages	8
ISBN (Print)	978-1-948087-81-0
Publication status	Published - 19 Nov 2020
Event	Fifth Conference on Machine Translation - Online Conference Duration: 19 Nov 2020 → 20 Nov 2020 http://www.statmt.org/wmt20/

Conference

Conference	Fifth Conference on Machine Translation
Abbreviated title	WMT 2020
City	Online Conference
Period	19/11/20 → 20/11/20
Internet address	http://www.statmt.org/wmt20/

Keywords / Materials (for Non-textual outputs)

machine translation
WMT
shared task
Tamil
Inuktitut

Access to Document

UEDIN_WMT20_System_descriptionAccepted author manuscript, 163 KBLicence: Creative Commons: Attribution (CC-BY)
The University of Edinburgh’s English-Tamil_BAWDEN_DOA29092020_VOR_CC-BYFinal published version, 234 KBLicence: Creative Commons: Attribution (CC-BY)

https://www.aclweb.org/anthology/2020.wmt-1.5Licence: Creative Commons: Attribution (CC-BY)

Global Under-Resourced MEdia Translation
Birch-Mayne, A. (Principal Investigator) & Haddow, B. (Co-investigator)
EU government bodies
1/01/19 → 30/06/22
Project: Research
MTStretch: Low-resource Machine Translation
Birch-Mayne, A. (Principal Investigator)
EPSRC
29/06/18 → 28/12/21
Project: Research

Cite this

Bawden, R., Birch, A., Dobreva, R., Oncevay Marcos, A., Miceli Barone, A. V., & Williams, P. (2020). The University of Edinburgh’s English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task. In Proceedings of the 5th Conference on Machine Translation (pp. 92-99). Association for Computational Linguistics (ACL). https://www.aclweb.org/anthology/2020.wmt-1.5

@inproceedings{2750d11429134cc191feb6dfe64beb83,

title = "The University of Edinburgh{\textquoteright}s English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task",

abstract = "We describe the University of Edinburgh's submissions to the WMT20 news translation shared task for the low resource language pair English-Tamil and the mid-resource language pair English-Inuktitut. We use the neural machine translation transformer architecture for all submissions and explore a variety of techniques to improve translation quality to compensate for the lack of parallel training data. For the very low-resource English-Tamil, this involves exploring pretraining, using both language model objectives and translation using an unrelated high-resource language pair (German-English), and iterative backtranslation. For English-Inuktitut, we explore the use of multilingual systems, which, despite not being part of the primary submission, would have achieved the best results on the test set.",

keywords = "machine translation, WMT, shared task, Tamil, Inuktitut",

author = "Rachel Bawden and Alexandra Birch and Radina Dobreva and \{Oncevay Marcos\}, Arturo and \{Miceli Barone\}, \{Antonio Valerio\} and Philip Williams",

year = "2020",

month = nov,

day = "19",

language = "English",

isbn = "978-1-948087-81-0",

pages = "92--99",

booktitle = "Proceedings of the 5th Conference on Machine Translation",

publisher = "Association for Computational Linguistics (ACL)",

note = "Fifth Conference on Machine Translation, WMT 2020 ; Conference date: 19-11-2020 Through 20-11-2020",

url = "http://www.statmt.org/wmt20/",

}

Bawden, R, Birch, A, Dobreva, R, Oncevay Marcos, A, Miceli Barone, AV & Williams, P 2020, The University of Edinburgh’s English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task. in Proceedings of the 5th Conference on Machine Translation. Association for Computational Linguistics (ACL), pp. 92-99, Fifth Conference on Machine Translation, Online Conference, 19/11/20. <https://www.aclweb.org/anthology/2020.wmt-1.5>

The University of Edinburgh’s English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task. / Bawden, Rachel; Birch, Alexandra; Dobreva, Radina et al.
Proceedings of the 5th Conference on Machine Translation. Association for Computational Linguistics (ACL), 2020. p. 92-99.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

TY - GEN

T1 - The University of Edinburgh’s English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task

AU - Bawden, Rachel

AU - Birch, Alexandra

AU - Dobreva, Radina

AU - Oncevay Marcos, Arturo

AU - Miceli Barone, Antonio Valerio

AU - Williams, Philip

PY - 2020/11/19

Y1 - 2020/11/19

N2 - We describe the University of Edinburgh's submissions to the WMT20 news translation shared task for the low resource language pair English-Tamil and the mid-resource language pair English-Inuktitut. We use the neural machine translation transformer architecture for all submissions and explore a variety of techniques to improve translation quality to compensate for the lack of parallel training data. For the very low-resource English-Tamil, this involves exploring pretraining, using both language model objectives and translation using an unrelated high-resource language pair (German-English), and iterative backtranslation. For English-Inuktitut, we explore the use of multilingual systems, which, despite not being part of the primary submission, would have achieved the best results on the test set.

AB - We describe the University of Edinburgh's submissions to the WMT20 news translation shared task for the low resource language pair English-Tamil and the mid-resource language pair English-Inuktitut. We use the neural machine translation transformer architecture for all submissions and explore a variety of techniques to improve translation quality to compensate for the lack of parallel training data. For the very low-resource English-Tamil, this involves exploring pretraining, using both language model objectives and translation using an unrelated high-resource language pair (German-English), and iterative backtranslation. For English-Inuktitut, we explore the use of multilingual systems, which, despite not being part of the primary submission, would have achieved the best results on the test set.

KW - machine translation

KW - WMT

KW - shared task

KW - Tamil

KW - Inuktitut

M3 - Conference contribution

SN - 978-1-948087-81-0

SP - 92

EP - 99

BT - Proceedings of the 5th Conference on Machine Translation

PB - Association for Computational Linguistics (ACL)

T2 - Fifth Conference on Machine Translation

Y2 - 19 November 2020 through 20 November 2020

ER -

The University of Edinburgh’s English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task

Abstract

Conference

Keywords / Materials (for Non-textual outputs)

Access to Document

Fingerprint

Projects

Global Under-Resourced MEdia Translation

MTStretch: Low-resource Machine Translation

Cite this