Four Approaches to Low-Resource Multilingual NMT: The Helsinki Submission to the AmericasNLP 2023 Shared Task

Ona De Gibert Bonet; Raúl Vázquez; Mikko Aulamo; Yves Scherrer; Sami Virpioja; Jörg Tiedemann

doi:10.18653/v1/2023.americasnlp-1.20

Four Approaches to Low-Resource Multilingual NMT: The Helsinki Submission to the AmericasNLP 2023 Shared Task

Ona De Gibert Bonet, Raúl Vázquez, Mikko Aulamo, Yves Scherrer, Sami Virpioja, Jörg Tiedemann

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Scientific › peer-review

Abstract

The Helsinki-NLP team participated in the AmericasNLP 2023 Shared Task with 6 submissions for all 11 language pairs arising from 4 different multilingual systems. We provide a detailed look at the work that went into collecting and preprocessing the data that led to our submissions. We explore various setups for multilingual Neural Machine Translation (NMT), namely knowledge distillation and transfer learning, multilingual NMT including a high-resource language (English), language-specific fine-tuning, and multilingual NMT exclusively using low-resource data. Our multilingual Model B ranks first in 4 out of the 11 language pairs.

Original language	English
Title of host publication	Proceedings of the Workshop on Natural Language Processing for Indigenous Languages of the Americas (AmericasNLP)
Editors	Manuel Mager, Abteen Ebrahimi, Arturo Oncevay, et al.
Number of pages	15
Place of Publication	Stroudsburg
Publisher	The Association for Computational Linguistics
Publication date	1 Jul 2023
Pages	177-191
ISBN (Electronic)	978-1-959429-91-3
DOIs	https://doi.org/10.18653/v1/2023.americasnlp-1.20
Publication status	Published - 1 Jul 2023
MoE publication type	A4 Article in conference proceedings
Event	Workshop on Natural Language Processing for indigenous Languages of the Americas - Toronto, Canada Duration: 14 Jun 2023 → 14 Jun 2023 Conference number: 3 https://turing.iimas.unam.mx/americasnlp/2023_workshop.html

Fields of Science

6121 Languages

Access to Document

10.18653/v1/2023.americasnlp-1.20Licence: CC BY

2023.americasnlp-1.20Final published version, 243 KBLicence: CC BY

1 Finished
1 Active

High Performance Language Technologies
Tiedemann, J. (Project manager), Aulamo, M. (Participant), De Gibert Bonet, O. (Participant), Grönroos, S.-A. (Participant), Ji, S. (Participant) & Virpioja, S. P. (Participant)
Charles University in Prague Faculty of Science Department of Teaching and Didactics of Biology
01/09/2022 → 31/08/2025
Project: EU Horizon Europe: Innovation actions (HORIZON-IA)
FoTran: Found in Translation - Natural Language Understanding with Cross-Lingual Grounding
Tiedemann, J. (Project manager), Attieh, J. (Participant), Aulamo, M. (Participant), Celikkanat, H. (Participant), De Gibert Bonet, O. (Participant), Grönroos, S.-A. (Participant), Mickus, T. (Participant), Raganato, A. (Participant), Scherrer, Y. (Participant), Silfverberg, M. (Participant), Sjöblom, E. I. (Participant), Talman, A. (Participant), Vazquez , R. (Participant), Virpioja, S. P. (Participant), Yli-Jyrä, A. (Participant), Zosa, E. (Participant), Celikkanat, H. (Participant), Raganato, A. (Participant), Silfverberg, M. (Participant), Sulubacak, U. (Participant) & Vazquez , R. (Participant)
01/09/2018 → 31/03/2024
Project: EU Horizon 2020: European Research Council: Consolidator Grant (H2020-ERC-COG)

Cite this

De Gibert Bonet, O., Vázquez, R., Aulamo, M., Scherrer, Y., Virpioja, S., & Tiedemann, J. (2023). Four Approaches to Low-Resource Multilingual NMT: The Helsinki Submission to the AmericasNLP 2023 Shared Task. In M. Mager, A. Ebrahimi, & A. Oncevay, et al. (Eds.), Proceedings of the Workshop on Natural Language Processing for Indigenous Languages of the Americas (AmericasNLP) (pp. 177-191). The Association for Computational Linguistics. https://doi.org/10.18653/v1/2023.americasnlp-1.20

De Gibert Bonet, Ona ; Vázquez, Raúl ; Aulamo, Mikko et al. / Four Approaches to Low-Resource Multilingual NMT : The Helsinki Submission to the AmericasNLP 2023 Shared Task. Proceedings of the Workshop on Natural Language Processing for Indigenous Languages of the Americas (AmericasNLP). editor / Manuel Mager ; Abteen Ebrahimi ; Arturo Oncevay, et al. Stroudsburg : The Association for Computational Linguistics, 2023. pp. 177-191

@inproceedings{d27c253a0cef43c1a10876343f0b4c87,

title = "Four Approaches to Low-Resource Multilingual NMT: The Helsinki Submission to the AmericasNLP 2023 Shared Task",

abstract = "The Helsinki-NLP team participated in the AmericasNLP 2023 Shared Task with 6 submissions for all 11 language pairs arising from 4 different multilingual systems. We provide a detailed look at the work that went into collecting and preprocessing the data that led to our submissions. We explore various setups for multilingual Neural Machine Translation (NMT), namely knowledge distillation and transfer learning, multilingual NMT including a high-resource language (English), language-specific fine-tuning, and multilingual NMT exclusively using low-resource data. Our multilingual Model B ranks first in 4 out of the 11 language pairs.",

keywords = "6121 Languages",

author = "{De Gibert Bonet}, Ona and Ra{\'u}l V{\'a}zquez and Mikko Aulamo and Yves Scherrer and Sami Virpioja and J{\"o}rg Tiedemann",

year = "2023",

month = jul,

day = "1",

doi = "10.18653/v1/2023.americasnlp-1.20",

language = "English",

pages = "177--191",

editor = "Mager, {Manuel } and Ebrahimi, {Abteen } and {Oncevay, et al.}, {Arturo }",

booktitle = "Proceedings of the Workshop on Natural Language Processing for Indigenous Languages of the Americas (AmericasNLP)",

publisher = "The Association for Computational Linguistics",

address = "United States",

note = "Workshop on Natural Language Processing for indigenous Languages of the Americas ; Conference date: 14-06-2023 Through 14-06-2023",

url = "https://turing.iimas.unam.mx/americasnlp/2023_workshop.html",

}

De Gibert Bonet, O , Vázquez, R , Aulamo, M , Scherrer, Y , Virpioja, S & Tiedemann, J 2023, Four Approaches to Low-Resource Multilingual NMT: The Helsinki Submission to the AmericasNLP 2023 Shared Task. in M Mager, A Ebrahimi & A Oncevay, et al. (eds), Proceedings of the Workshop on Natural Language Processing for Indigenous Languages of the Americas (AmericasNLP). The Association for Computational Linguistics, Stroudsburg, pp. 177-191, Workshop on Natural Language Processing for indigenous Languages of the Americas, Toronto, Canada, 14/06/2023. https://doi.org/10.18653/v1/2023.americasnlp-1.20

Four Approaches to Low-Resource Multilingual NMT: The Helsinki Submission to the AmericasNLP 2023 Shared Task. / De Gibert Bonet, Ona ; Vázquez, Raúl ; Aulamo, Mikko et al.
Proceedings of the Workshop on Natural Language Processing for Indigenous Languages of the Americas (AmericasNLP). ed. / Manuel Mager; Abteen Ebrahimi; Arturo Oncevay, et al. Stroudsburg: The Association for Computational Linguistics, 2023. p. 177-191.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Scientific › peer-review

TY - GEN

T1 - Four Approaches to Low-Resource Multilingual NMT

T2 - Workshop on Natural Language Processing for indigenous Languages of the Americas

AU - De Gibert Bonet, Ona

AU - Vázquez, Raúl

AU - Aulamo, Mikko

AU - Scherrer, Yves

AU - Virpioja, Sami

AU - Tiedemann, Jörg

N1 - Conference code: 3

PY - 2023/7/1

Y1 - 2023/7/1

N2 - The Helsinki-NLP team participated in the AmericasNLP 2023 Shared Task with 6 submissions for all 11 language pairs arising from 4 different multilingual systems. We provide a detailed look at the work that went into collecting and preprocessing the data that led to our submissions. We explore various setups for multilingual Neural Machine Translation (NMT), namely knowledge distillation and transfer learning, multilingual NMT including a high-resource language (English), language-specific fine-tuning, and multilingual NMT exclusively using low-resource data. Our multilingual Model B ranks first in 4 out of the 11 language pairs.

AB - The Helsinki-NLP team participated in the AmericasNLP 2023 Shared Task with 6 submissions for all 11 language pairs arising from 4 different multilingual systems. We provide a detailed look at the work that went into collecting and preprocessing the data that led to our submissions. We explore various setups for multilingual Neural Machine Translation (NMT), namely knowledge distillation and transfer learning, multilingual NMT including a high-resource language (English), language-specific fine-tuning, and multilingual NMT exclusively using low-resource data. Our multilingual Model B ranks first in 4 out of the 11 language pairs.

KW - 6121 Languages

U2 - 10.18653/v1/2023.americasnlp-1.20

DO - 10.18653/v1/2023.americasnlp-1.20

M3 - Conference contribution

SP - 177

EP - 191

BT - Proceedings of the Workshop on Natural Language Processing for Indigenous Languages of the Americas (AmericasNLP)

A2 - Mager, Manuel

A2 - Ebrahimi, Abteen

A2 - Oncevay, et al., Arturo

PB - The Association for Computational Linguistics

CY - Stroudsburg

Y2 - 14 June 2023 through 14 June 2023

ER -

De Gibert Bonet O , Vázquez R , Aulamo M , Scherrer Y , Virpioja S , Tiedemann J. Four Approaches to Low-Resource Multilingual NMT: The Helsinki Submission to the AmericasNLP 2023 Shared Task. In Mager M, Ebrahimi A, Oncevay, et al. A, editors, Proceedings of the Workshop on Natural Language Processing for Indigenous Languages of the Americas (AmericasNLP). Stroudsburg: The Association for Computational Linguistics. 2023. p. 177-191 doi: 10.18653/v1/2023.americasnlp-1.20

Four Approaches to Low-Resource Multilingual NMT: The Helsinki Submission to the AmericasNLP 2023 Shared Task

Abstract

Fields of Science

Access to Document

Projects

High Performance Language Technologies

FoTran: Found in Translation - Natural Language Understanding with Cross-Lingual Grounding

Cite this