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

Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review

84 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 languageEnglish
Title of host publicationSecond Conference on Machine Translation (WMT17); Copenhagen, Denmark
Pages296–302
Number of pages7
ISBN (Electronic)978-1-945626-01-2
Publication statusPublished - 7 Sep 2017
MoE publication typeA4 Article in a conference publication
EventConference on Machine Translation - Copenhagen, Denmark, Copenhagen, Denmark
Duration: 7 Sep 201711 Sep 2017
Conference number: 2

Conference

ConferenceConference on Machine Translation
Abbreviated titleWMT
CountryDenmark
CityCopenhagen
Period07/09/201711/09/2017

Keywords

  • neural machine translation
  • morphology
  • multi-task learning

Equipment

Science-IT

Mikko Hakala (Manager)

School of Science

Facility/equipment: Facility

  • Cite this