Empirically sampling Universal Dependencies

Natalie Schluter, Zeljko Agic

Publikation: Artikel i tidsskrift og konference artikel i tidsskriftKonferenceartikelForskningpeer review

Abstract

Universal Dependencies incur a high cost in computation for unbiased system development. We propose a 100% empirically chosen small subset of UD languages for efficient parsing system development. The technique used is based on measurements of model capacity globally. We show that the diversity of the resulting representative language set is superior to the requirements-based procedure.
OriginalsprogEngelsk
TidsskriftNEALT (Northern European Association of Language Technology) Proceedings Series
Vol/bind31
Sider (fra-til)117-122
ISSN1736-6305
StatusUdgivet - 2017
BegivenhedNoDaLiDa 2017 Workshop on Universal Dependencies - Gothenburg, Sverige
Varighed: 22 maj 2017 → …

Workshop

WorkshopNoDaLiDa 2017 Workshop on Universal Dependencies
Land/OmrådeSverige
ByGothenburg
Periode22/05/2017 → …

Emneord

  • Universal Dependencies
  • Parsers
  • Language subset selection
  • Model capacity
  • Empirical methods

Fingeraftryk

Dyk ned i forskningsemnerne om 'Empirically sampling Universal Dependencies'. Sammen danner de et unikt fingeraftryk.

Citationsformater