Extending hybrid word-character neural machine translation with multi-task learning of morphological analysis

Stig-Arne Grönroos; Sami Virpioja; Mikko Kurimo

Extending hybrid word-character neural machine translation with multi-task learning of morphological analysis

Stig-Arne Grönroos
, Sami Virpioja
, Mikko Kurimo

Speech Recognition

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

8 Citations (Scopus)

131 Downloads (Pure)

Abstract

This article describes the Aalto University entry to the English-to-Finnish news translation shared task in WMT 2017. Our system is an open vocabulary neural machine translation (NMT) system, adapted to the needs of a morphologically complex target language. The main contributions of this paper are 1) implicitly incorporating morphological information to NMT through multi-task learning, 2) adding an attention mechanism to the character-level decoder, combined with character segmentation of names, and 3) a new overattending penalty to beam search.

Original language	English
Title of host publication	Second Conference on Machine Translation (WMT17); Copenhagen, Denmark
Publisher	Association for Computational Linguistics
Pages	296–302
Number of pages	7
ISBN (Electronic)	978-1-945626-01-2
Publication status	Published - 7 Sept 2017
MoE publication type	A4 Conference publication
Event	Conference on Machine Translation - Copenhagen, Denmark, Copenhagen, Denmark Duration: 7 Sept 2017 → 11 Sept 2017 Conference number: 2

Conference

Conference	Conference on Machine Translation
Abbreviated title	WMT
Country/Territory	Denmark
City	Copenhagen
Period	07/09/2017 → 11/09/2017

Keywords

neural machine translation
morphology
multi-task learning

Access to Document

gronroos-et-al-WMTFinal published version, 120 KBLicence: CC BY

http://www.aclweb.org/anthology/W17-4727Licence: CC BY

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

Cite this

@inproceedings{e73707488a6a484894326daa9410f870,

title = "Extending hybrid word-character neural machine translation with multi-task learning of morphological analysis",

abstract = "This article describes the Aalto University entry to the English-to-Finnish news translation shared task in WMT 2017. Our system is an open vocabulary neural machine translation (NMT) system, adapted to the needs of a morphologically complex target language. The main contributions of this paper are 1) implicitly incorporating morphological information to NMT through multi-task learning, 2) adding an attention mechanism to the character-level decoder, combined with character segmentation of names, and 3) a new overattending penalty to beam search.",

keywords = "neural machine translation, morphology, multi-task learning",

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

note = "VK: COIN; Conference on Machine Translation, WMT ; Conference date: 07-09-2017 Through 11-09-2017",

year = "2017",

month = sep,

day = "7",

language = "English",

pages = "296–302",

booktitle = "Second Conference on Machine Translation (WMT17); Copenhagen, Denmark",

publisher = "Association for Computational Linguistics",

address = "United States",

}

Grönroos, S-A, Virpioja, S & Kurimo, M 2017, Extending hybrid word-character neural machine translation with multi-task learning of morphological analysis. in Second Conference on Machine Translation (WMT17); Copenhagen, Denmark. Association for Computational Linguistics, pp. 296–302, Conference on Machine Translation, Copenhagen, Denmark, 07/09/2017. <http://www.aclweb.org/anthology/W17-4727>

Extending hybrid word-character neural machine translation with multi-task learning of morphological analysis. / Grönroos, Stig-Arne; Virpioja, Sami; Kurimo, Mikko.
Second Conference on Machine Translation (WMT17); Copenhagen, Denmark. Association for Computational Linguistics, 2017. p. 296–302.

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

TY - GEN

T1 - Extending hybrid word-character neural machine translation with multi-task learning of morphological analysis

AU - Grönroos, Stig-Arne

AU - Virpioja, Sami

AU - Kurimo, Mikko

N1 - Conference code: 2

PY - 2017/9/7

Y1 - 2017/9/7

N2 - This article describes the Aalto University entry to the English-to-Finnish news translation shared task in WMT 2017. Our system is an open vocabulary neural machine translation (NMT) system, adapted to the needs of a morphologically complex target language. The main contributions of this paper are 1) implicitly incorporating morphological information to NMT through multi-task learning, 2) adding an attention mechanism to the character-level decoder, combined with character segmentation of names, and 3) a new overattending penalty to beam search.

AB - This article describes the Aalto University entry to the English-to-Finnish news translation shared task in WMT 2017. Our system is an open vocabulary neural machine translation (NMT) system, adapted to the needs of a morphologically complex target language. The main contributions of this paper are 1) implicitly incorporating morphological information to NMT through multi-task learning, 2) adding an attention mechanism to the character-level decoder, combined with character segmentation of names, and 3) a new overattending penalty to beam search.

KW - neural machine translation

KW - morphology

KW - multi-task learning

M3 - Conference article in proceedings

SP - 296

EP - 302

BT - Second Conference on Machine Translation (WMT17); Copenhagen, Denmark

PB - Association for Computational Linguistics

T2 - Conference on Machine Translation

Y2 - 7 September 2017 through 11 September 2017

ER -

Extending hybrid word-character neural machine translation with multi-task learning of morphological analysis

Abstract

Conference

Keywords

Access to Document

Fingerprint

Equipment

Science-IT

Cite this