Cognate-aware morphological segmentation for multilingual neural translation

Stig-Arne Grönroos; Sami Virpioja; Mikko Kurimo

doi:10.18653/v1/W18-64037

Cognate-aware morphological segmentation for multilingual neural translation

Stig-Arne Grönroos, Sami Virpioja, Mikko Kurimo

Speech Recognition

Utopia Analytics Oy

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

10 Citations (Scopus)

134 Downloads (Pure)

Abstract

This article describes the Aalto University entry to the WMT18 News Translation Shared Task. We participate in the multilingual subtrack with a system trained under the constrained condition to translate from English to both Finnish and Estonian. The system is based on the Transformer model. We focus on improving the consistency of morphological segmentation for words that are similar orthographically, semantically, and distributionally; such words include etymological cognates, loan words, and proper names. For this, we introduce Cognate Morfessor, a multilingual variant of the Morfessor method. We show that our approach improves the translation quality particularly for Estonian, which has less resources for training the translation model.

Original language	English
Title of host publication	Third Conference on Machine Translation (WMT18); Brussels, Belgium
Publisher	Association for Computational Linguistics
Pages	390-397
Number of pages	8
ISBN (Print)	978-1-948087-81-0
DOIs	https://doi.org/10.18653/v1/W18-64037
Publication status	Published - 31 Oct 2018
MoE publication type	A4 Conference publication
Event	Conference on Machine Translation - Brussels, Belgium Duration: 31 Oct 2018 → 1 Dec 2018 Conference number: 3

Conference

Conference	Conference on Machine Translation
Abbreviated title	WMT
Country/Territory	Belgium
City	Brussels
Period	31/10/2018 → 01/12/2018

Keywords

neural machine translation
morphology
cognate
multilingual

Access to Document

10.18653/v1/W18-64037Licence: CC BY

WMT037(1)Final published version, 416 KBLicence: CC BY

http://aclweb.org/anthology/W18-6410Licence: CC BY

MeMAD: Methods for Managing Audiovisual Data: Combining Automatic Efficiency with Human Accuracy
Kurimo, M. (Principal investigator), Grönroos, S.-A. (Project Member), Brander, T. (Project Member), Porjazovski, D. (Project Member), Raitio, R. (Project Member), Grósz, T. (Project Member), Virkkunen, A. (Project Member) & Rouhe, A. (Project Member)
27/12/2017 → 31/03/2021
Project: EU: Framework programmes funding

Science-IT
Hakala, M. (Manager)
School of Science
Facility/equipment: Facility

Cite this

@inproceedings{66d66e8f6c09476384ad532d0e9809e2,

title = "Cognate-aware morphological segmentation for multilingual neural translation",

abstract = "This article describes the Aalto University entry to the WMT18 News Translation Shared Task. We participate in the multilingual subtrack with a system trained under the constrained condition to translate from English to both Finnish and Estonian. The system is based on the Transformer model. We focus on improving the consistency of morphological segmentation for words that are similar orthographically, semantically, and distributionally; such words include etymological cognates, loan words, and proper names. For this, we introduce Cognate Morfessor, a multilingual variant of the Morfessor method. We show that our approach improves the translation quality particularly for Estonian, which has less resources for training the translation model.",

keywords = "neural machine translation, morphology, cognate, multilingual",

author = "Stig-Arne Gr{\"o}nroos and Sami Virpioja and Mikko Kurimo",

note = "| openaire: EC/H2020/780069/EU//MeMAD; Conference on Machine Translation, WMT ; Conference date: 31-10-2018 Through 01-12-2018",

year = "2018",

month = oct,

day = "31",

doi = "10.18653/v1/W18-64037",

language = "English",

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

pages = "390--397",

booktitle = "Third Conference on Machine Translation (WMT18); Brussels, Belgium",

publisher = "Association for Computational Linguistics",

address = "United States",

}

Cognate-aware morphological segmentation for multilingual neural translation. / Grönroos, Stig-Arne; Virpioja, Sami; Kurimo, Mikko.
Third Conference on Machine Translation (WMT18); Brussels, Belgium. Association for Computational Linguistics, 2018. p. 390-397.

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

TY - GEN

T1 - Cognate-aware morphological segmentation for multilingual neural translation

AU - Grönroos, Stig-Arne

AU - Virpioja, Sami

AU - Kurimo, Mikko

N1 - Conference code: 3

PY - 2018/10/31

Y1 - 2018/10/31

N2 - This article describes the Aalto University entry to the WMT18 News Translation Shared Task. We participate in the multilingual subtrack with a system trained under the constrained condition to translate from English to both Finnish and Estonian. The system is based on the Transformer model. We focus on improving the consistency of morphological segmentation for words that are similar orthographically, semantically, and distributionally; such words include etymological cognates, loan words, and proper names. For this, we introduce Cognate Morfessor, a multilingual variant of the Morfessor method. We show that our approach improves the translation quality particularly for Estonian, which has less resources for training the translation model.

AB - This article describes the Aalto University entry to the WMT18 News Translation Shared Task. We participate in the multilingual subtrack with a system trained under the constrained condition to translate from English to both Finnish and Estonian. The system is based on the Transformer model. We focus on improving the consistency of morphological segmentation for words that are similar orthographically, semantically, and distributionally; such words include etymological cognates, loan words, and proper names. For this, we introduce Cognate Morfessor, a multilingual variant of the Morfessor method. We show that our approach improves the translation quality particularly for Estonian, which has less resources for training the translation model.

KW - neural machine translation

KW - morphology

KW - cognate

KW - multilingual

U2 - 10.18653/v1/W18-64037

DO - 10.18653/v1/W18-64037

M3 - Conference article in proceedings

SN - 978-1-948087-81-0

SP - 390

EP - 397

BT - Third Conference on Machine Translation (WMT18); Brussels, Belgium

PB - Association for Computational Linguistics

T2 - Conference on Machine Translation

Y2 - 31 October 2018 through 1 December 2018

ER -

Cognate-aware morphological segmentation for multilingual neural translation

Abstract

Conference

Keywords

Access to Document

Fingerprint

Projects

MeMAD: Methods for Managing Audiovisual Data: Combining Automatic Efficiency with Human Accuracy

Equipment

Science-IT

Cite this