The University of Edinburgh's English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task
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 vari...
Main Authors: | , , , , , |
---|---|
Format: | Conference Object |
Language: | unknown |
Published: |
Zenodo
2020
|
Subjects: | |
Online Access: | https://doi.org/10.5281/zenodo.6672692 |
id |
ftzenodo:oai:zenodo.org:6672692 |
---|---|
record_format |
openpolar |
spelling |
ftzenodo:oai:zenodo.org:6672692 2024-09-15T18:15:08+00:00 The University of Edinburgh's English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task Bawden Birch Dobreva Oncevay Miceli Williams 2020-01-01 https://doi.org/10.5281/zenodo.6672692 unknown Zenodo https://zenodo.org/communities/eu https://doi.org/10.5281/zenodo.6672691 https://doi.org/10.5281/zenodo.6672692 oai:zenodo.org:6672692 info:eu-repo/semantics/openAccess Creative Commons Attribution 4.0 International https://creativecommons.org/licenses/by/4.0/legalcode Conference on Machine Translation (WMT) info:eu-repo/semantics/conferencePaper 2020 ftzenodo https://doi.org/10.5281/zenodo.667269210.5281/zenodo.6672691 2024-07-27T05:32:11Z 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. Conference Object inuktitut Zenodo |
institution |
Open Polar |
collection |
Zenodo |
op_collection_id |
ftzenodo |
language |
unknown |
description |
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. |
format |
Conference Object |
author |
Bawden Birch Dobreva Oncevay Miceli Williams |
spellingShingle |
Bawden Birch Dobreva Oncevay Miceli Williams The University of Edinburgh's English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task |
author_facet |
Bawden Birch Dobreva Oncevay Miceli Williams |
author_sort |
Bawden |
title |
The University of Edinburgh's English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task |
title_short |
The University of Edinburgh's English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task |
title_full |
The University of Edinburgh's English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task |
title_fullStr |
The University of Edinburgh's English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task |
title_full_unstemmed |
The University of Edinburgh's English-Tamil and English-Inuktitut Submissions to the WMT20 News Translation Task |
title_sort |
university of edinburgh's english-tamil and english-inuktitut submissions to the wmt20 news translation task |
publisher |
Zenodo |
publishDate |
2020 |
url |
https://doi.org/10.5281/zenodo.6672692 |
genre |
inuktitut |
genre_facet |
inuktitut |
op_source |
Conference on Machine Translation (WMT) |
op_relation |
https://zenodo.org/communities/eu https://doi.org/10.5281/zenodo.6672691 https://doi.org/10.5281/zenodo.6672692 oai:zenodo.org:6672692 |
op_rights |
info:eu-repo/semantics/openAccess Creative Commons Attribution 4.0 International https://creativecommons.org/licenses/by/4.0/legalcode |
op_doi |
https://doi.org/10.5281/zenodo.667269210.5281/zenodo.6672691 |
_version_ |
1810452866310078464 |