Evaluating the morphological competence of Machine Translation Systems

Abstract : While recent changes in Machine Translation state-of-the-art brought translation quality a step further, it is regularly acknowledged that the standard automatic metrics do not provide enough insights to fully measure the impact of neural models. This paper proposes a new type of evaluation focused specifically on the morphological competence of a system with respect to various grammatical phenomena. Our approach uses automatically generated pairs of source sentences, where each pair tests one morphological contrast. This methodology is used to compare several systems submitted at WMT'17 for English into Czech and Latvian.
Type de document :
Communication dans un congrès
Association for Computational Linguistics. 2nd Conference on Machine Translation (WMT17), Sep 2017, Copenhague, Denmark. Proceedings of the Conference on Machine Translation (WMT), 1, pp.43-55, 2017, 〈http://www.statmt.org/wmt17/〉
Liste complète des métadonnées

Littérature citée [36 références]  Voir  Masquer  Télécharger

https://hal.archives-ouvertes.fr/hal-01618387
Contributeur : Dev.Limsi Dev.Limsi <>
Soumis le : jeudi 19 octobre 2017 - 14:43:06
Dernière modification le : jeudi 11 janvier 2018 - 06:25:47
Document(s) archivé(s) le : samedi 20 janvier 2018 - 12:24:42

Fichier

W17-4705.pdf
Fichiers éditeurs autorisés sur une archive ouverte

Identifiants

  • HAL Id : hal-01618387, version 1

Collections

Citation

Franck Burlot, François Yvon. Evaluating the morphological competence of Machine Translation Systems. Association for Computational Linguistics. 2nd Conference on Machine Translation (WMT17), Sep 2017, Copenhague, Denmark. Proceedings of the Conference on Machine Translation (WMT), 1, pp.43-55, 2017, 〈http://www.statmt.org/wmt17/〉. 〈hal-01618387〉

Partager

Métriques

Consultations de la notice

63

Téléchargements de fichiers

24