A System for Multilingual Dependency Parsing based on Bidirectional LSTM Feature Representations

Abstract : In this paper, we present our multilingual dependency parser developed for the CoNLL 2017 UD Shared Task dealing with " Multilingual Parsing from Raw Text to Universal Dependencies " 1. Our parser extends the monolingual BIST-parser as a multi-source multilingual trainable parser. Thanks to multilingual word embeddings and one hot encodings for languages, our system can use both monolingual and multi-source training. We trained 69 monolingual language models and 13 multilingual models for the shared task. Our multilingual approach making use of different resources yield better results than the monolingual approach for 11 languages. Our system ranked 5 th and achieved 70.93 overall LAS score over the 81 test corpora (macro-averaged LAS F1 score).
Type de document :
Communication dans un congrès
Computational Natural Language Learning (CoNLL), Aug 2017, Vancouver, Canada. Proceedings of the CoNLL 2017 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies pp.63 - 70, 〈https://aclanthology.coli.uni-saarland.de/papers/K17-3006/k17-3006〉
Liste complète des métadonnées

Littérature citée [17 références]  Voir  Masquer  Télécharger

https://hal.archives-ouvertes.fr/hal-01722370
Contributeur : Thierry Poibeau <>
Soumis le : lundi 12 mars 2018 - 00:24:58
Dernière modification le : mardi 24 avril 2018 - 17:20:13
Document(s) archivé(s) le : mercredi 13 juin 2018 - 12:36:34

Fichier

K17-3006.pdf
Fichiers éditeurs autorisés sur une archive ouverte

Identifiants

  • HAL Id : hal-01722370, version 1

Collections

Citation

Kyungtae Lim, Thierry Poibeau. A System for Multilingual Dependency Parsing based on Bidirectional LSTM Feature Representations. Computational Natural Language Learning (CoNLL), Aug 2017, Vancouver, Canada. Proceedings of the CoNLL 2017 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies pp.63 - 70, 〈https://aclanthology.coli.uni-saarland.de/papers/K17-3006/k17-3006〉. 〈hal-01722370〉

Partager

Métriques

Consultations de la notice

46

Téléchargements de fichiers

21