Maastricht University’s Multilingual Speech Translation System for IWSLT 2021

Danni Liu; Jan Niehues

doi:10.18653/v1/2021.iwslt-1.15

Maastricht University’s Multilingual Speech Translation System for IWSLT 2021

Danni Liu^*, Jan Niehues

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceeding › Academic › peer-review

Abstract

This paper describes Maastricht University’s participation in the IWSLT 2021 multilingual speech translation track. The task in this track is to build multilingual speech translation systems in supervised and zero-shot directions. Our primary system is an end-to-end model that performs both speech transcription and translation. We observe that the joint training for the two tasks is complementary especially when the speech translation data is scarce. On the source and target side, we use data augmentation and pseudo-labels respectively to improve the performance of our systems. We also introduce an ensembling technique that consistently improves the quality of transcriptions and translations. The experiments show that the end-to-end system is competitive with its cascaded counterpart especially in zero-shot conditions.

Original language	English
Title of host publication	Proceedings of the 18th International Conference on Spoken Language Translation (IWSLT 2021)
Editors	Marcello Federico, Alex Waibel, Marta R. Costa-jussa, Jan Niehues, Sebastian Stuker, Elizabeth Salesky
Publisher	Association for Computational Linguistics
Pages	138–143
Number of pages	6
Edition	August 2021
ISBN (Print)	9781954085749
DOIs	https://doi.org/10.18653/v1/2021.iwslt-1.15
Publication status	Published - 2021
Event	18th International Conference on Spoken Language Translation - Online, Bangkok, Thailand Duration: 5 Aug 2021 → 6 Aug 2021 https://iwslt.org/2021/

Conference

Conference	18th International Conference on Spoken Language Translation
Abbreviated title	IWSLT 2021
Country/Territory	Thailand
City	Bangkok
Period	5/08/21 → 6/08/21
Internet address	https://iwslt.org/2021/

Access to Document

10.18653/v1/2021.iwslt-1.15Licence: CC BY

https://aclanthology.org/2021.iwslt-1.15/

Cite this

Liu, D., & Niehues, J. (2021). Maastricht University’s Multilingual Speech Translation System for IWSLT 2021. In M. Federico, A. Waibel, M. R. Costa-jussa, J. Niehues, S. Stuker, & E. Salesky (Eds.), Proceedings of the 18th International Conference on Spoken Language Translation (IWSLT 2021) (August 2021 ed., pp. 138–143). Association for Computational Linguistics. https://doi.org/10.18653/v1/2021.iwslt-1.15

Liu, Danni ; Niehues, Jan. / Maastricht University’s Multilingual Speech Translation System for IWSLT 2021. Proceedings of the 18th International Conference on Spoken Language Translation (IWSLT 2021). editor / Marcello Federico ; Alex Waibel ; Marta R. Costa-jussa ; Jan Niehues ; Sebastian Stuker ; Elizabeth Salesky. August 2021. ed. Association for Computational Linguistics, 2021. pp. 138–143

@inproceedings{706c9c8efbad464bb9ffd859ab702847,

title = "Maastricht University{\textquoteright}s Multilingual Speech Translation System for IWSLT 2021",

abstract = "This paper describes Maastricht University{\textquoteright}s participation in the IWSLT 2021 multilingual speech translation track. The task in this track is to build multilingual speech translation systems in supervised and zero-shot directions. Our primary system is an end-to-end model that performs both speech transcription and translation. We observe that the joint training for the two tasks is complementary especially when the speech translation data is scarce. On the source and target side, we use data augmentation and pseudo-labels respectively to improve the performance of our systems. We also introduce an ensembling technique that consistently improves the quality of transcriptions and translations. The experiments show that the end-to-end system is competitive with its cascaded counterpart especially in zero-shot conditions.",

author = "Danni Liu and Jan Niehues",

note = "Funding Information: We thank the anonymous reviewers for their helpful feedback. This work is supported by a Facebook Sponsored Research Agreement. Publisher Copyright: {\textcopyright} 2021 Association for Computational Linguistics.; 18th International Conference on Spoken Language Translation, IWSLT 2021 ; Conference date: 05-08-2021 Through 06-08-2021",

year = "2021",

doi = "10.18653/v1/2021.iwslt-1.15",

language = "English",

isbn = "9781954085749",

pages = "138–143",

editor = "Marcello Federico and Alex Waibel and Costa-jussa, {Marta R.} and Jan Niehues and Sebastian Stuker and Elizabeth Salesky",

booktitle = "Proceedings of the 18th International Conference on Spoken Language Translation (IWSLT 2021)",

publisher = "Association for Computational Linguistics",

edition = "August 2021",

url = "https://iwslt.org/2021/",

}

Liu, D & Niehues, J 2021, Maastricht University’s Multilingual Speech Translation System for IWSLT 2021. in M Federico, A Waibel, MR Costa-jussa, J Niehues, S Stuker & E Salesky (eds), Proceedings of the 18th International Conference on Spoken Language Translation (IWSLT 2021). August 2021 edn, Association for Computational Linguistics, pp. 138–143, 18th International Conference on Spoken Language Translation, Bangkok, Thailand, 5/08/21. https://doi.org/10.18653/v1/2021.iwslt-1.15

Maastricht University’s Multilingual Speech Translation System for IWSLT 2021. / Liu, Danni; Niehues, Jan.
Proceedings of the 18th International Conference on Spoken Language Translation (IWSLT 2021). ed. / Marcello Federico; Alex Waibel; Marta R. Costa-jussa; Jan Niehues; Sebastian Stuker; Elizabeth Salesky. August 2021. ed. Association for Computational Linguistics, 2021. p. 138–143.

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceeding › Academic › peer-review

TY - GEN

T1 - Maastricht University’s Multilingual Speech Translation System for IWSLT 2021

AU - Liu, Danni

AU - Niehues, Jan

N1 - Funding Information: We thank the anonymous reviewers for their helpful feedback. This work is supported by a Facebook Sponsored Research Agreement. Publisher Copyright: © 2021 Association for Computational Linguistics.

PY - 2021

Y1 - 2021

N2 - This paper describes Maastricht University’s participation in the IWSLT 2021 multilingual speech translation track. The task in this track is to build multilingual speech translation systems in supervised and zero-shot directions. Our primary system is an end-to-end model that performs both speech transcription and translation. We observe that the joint training for the two tasks is complementary especially when the speech translation data is scarce. On the source and target side, we use data augmentation and pseudo-labels respectively to improve the performance of our systems. We also introduce an ensembling technique that consistently improves the quality of transcriptions and translations. The experiments show that the end-to-end system is competitive with its cascaded counterpart especially in zero-shot conditions.

AB - This paper describes Maastricht University’s participation in the IWSLT 2021 multilingual speech translation track. The task in this track is to build multilingual speech translation systems in supervised and zero-shot directions. Our primary system is an end-to-end model that performs both speech transcription and translation. We observe that the joint training for the two tasks is complementary especially when the speech translation data is scarce. On the source and target side, we use data augmentation and pseudo-labels respectively to improve the performance of our systems. We also introduce an ensembling technique that consistently improves the quality of transcriptions and translations. The experiments show that the end-to-end system is competitive with its cascaded counterpart especially in zero-shot conditions.

U2 - 10.18653/v1/2021.iwslt-1.15

DO - 10.18653/v1/2021.iwslt-1.15

M3 - Conference article in proceeding

SN - 9781954085749

SP - 138

EP - 143

BT - Proceedings of the 18th International Conference on Spoken Language Translation (IWSLT 2021)

A2 - Federico, Marcello

A2 - Waibel, Alex

A2 - Costa-jussa, Marta R.

A2 - Niehues, Jan

A2 - Stuker, Sebastian

A2 - Salesky, Elizabeth

PB - Association for Computational Linguistics

T2 - 18th International Conference on Spoken Language Translation

Y2 - 5 August 2021 through 6 August 2021

ER -

Liu D, Niehues J. Maastricht University’s Multilingual Speech Translation System for IWSLT 2021. In Federico M, Waibel A, Costa-jussa MR, Niehues J, Stuker S, Salesky E, editors, Proceedings of the 18th International Conference on Spoken Language Translation (IWSLT 2021). August 2021 ed. Association for Computational Linguistics. 2021. p. 138–143 doi: 10.18653/v1/2021.iwslt-1.15