TransQuest at WMT2020: Sentence-Level Direct Assessment

Tharindu Ranasinghe; Constantin Orasan; Ruslan Mitkov

TransQuest at WMT2020: Sentence-Level Direct Assessment

Tharindu Ranasinghe
, Constantin Orasan
, Ruslan Mitkov

University of Surrey
University of Wolverhampton

Research output: Chapter in Book/Published conference output › Conference publication

29 Citations (SciVal)

Abstract

This paper presents the team TransQuest’s participation in Sentence-Level Direct Assessment shared task in WMT 2020. We introduce a simple QE framework based on cross-lingual transformers, and we use it to implement and evaluate two different neural architectures. The proposed methods achieve state-of-the-art results surpassing the results obtained by OpenKiwi, the baseline used in the shared task. We further fine tune the QE framework by performing ensemble and data augmentation. Our approach is the winning solution in all of the language pairs according to the WMT 2020 official results.

Original language	English
Title of host publication	Fifth Conference on Machine Translation
Publisher	Association for Computational Linguistics (ACL)
Pages	1049–1055
Publication status	Published - Nov 2020
Event	5th Conference on Machine Translation - Duration: 19 Nov 2020 → 20 Nov 2020

Conference

Conference	5th Conference on Machine Translation
Abbreviated title	WMT 20
Period	19/11/20 → 20/11/20

Access to Document

https://aclanthology.org/2020.wmt-1.122

Cite this

@inproceedings{029299fd04004868a51128f68ef51c8f,

title = "TransQuest at WMT2020: Sentence-Level Direct Assessment",

abstract = "This paper presents the team TransQuest{\textquoteright}s participation in Sentence-Level Direct Assessment shared task in WMT 2020. We introduce a simple QE framework based on cross-lingual transformers, and we use it to implement and evaluate two different neural architectures. The proposed methods achieve state-of-the-art results surpassing the results obtained by OpenKiwi, the baseline used in the shared task. We further fine tune the QE framework by performing ensemble and data augmentation. Our approach is the winning solution in all of the language pairs according to the WMT 2020 official results.",

author = "Tharindu Ranasinghe and Constantin Orasan and Ruslan Mitkov",

year = "2020",

month = nov,

language = "English",

pages = "1049–1055",

booktitle = "Fifth Conference on Machine Translation",

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

address = "United States",

note = "5th Conference on Machine Translation, WMT 20 ; Conference date: 19-11-2020 Through 20-11-2020",

}

TY - GEN

T1 - TransQuest at WMT2020: Sentence-Level Direct Assessment

AU - Ranasinghe, Tharindu

AU - Orasan, Constantin

AU - Mitkov, Ruslan

PY - 2020/11

Y1 - 2020/11

N2 - This paper presents the team TransQuest’s participation in Sentence-Level Direct Assessment shared task in WMT 2020. We introduce a simple QE framework based on cross-lingual transformers, and we use it to implement and evaluate two different neural architectures. The proposed methods achieve state-of-the-art results surpassing the results obtained by OpenKiwi, the baseline used in the shared task. We further fine tune the QE framework by performing ensemble and data augmentation. Our approach is the winning solution in all of the language pairs according to the WMT 2020 official results.

AB - This paper presents the team TransQuest’s participation in Sentence-Level Direct Assessment shared task in WMT 2020. We introduce a simple QE framework based on cross-lingual transformers, and we use it to implement and evaluate two different neural architectures. The proposed methods achieve state-of-the-art results surpassing the results obtained by OpenKiwi, the baseline used in the shared task. We further fine tune the QE framework by performing ensemble and data augmentation. Our approach is the winning solution in all of the language pairs according to the WMT 2020 official results.

M3 - Conference publication

SP - 1049

EP - 1055

BT - Fifth Conference on Machine Translation

PB - Association for Computational Linguistics (ACL)

T2 - 5th Conference on Machine Translation

Y2 - 19 November 2020 through 20 November 2020

ER -

TransQuest at WMT2020: Sentence-Level Direct Assessment

Abstract

Conference

Access to Document

Fingerprint

Cite this