Scalable Cross-lingual Treebank Synthesis for Improved Production Dependency Parsers

Yousef El-Kurdi, Hiroshi Kanayama, Efsun Sarioglu Kayi, Vittorio Castelli, Todd Ward, Radu Florian


Abstract
We present scalable Universal Dependency (UD) treebank synthesis techniques that exploit advances in language representation modeling which leverage vast amounts of unlabeled general-purpose multilingual text. We introduce a data augmentation technique that uses synthetic treebanks to improve production-grade parsers. The synthetic treebanks are generated using a state-of-the-art biaffine parser adapted with pretrained Transformer models, such as Multilingual BERT (M-BERT). The new parser improves LAS by up to two points on seven languages. The production models’ LAS performance improves as the augmented treebanks scale in size, surpassing performance of production models trained on originally annotated UD treebanks.
Anthology ID:
2020.coling-industry.16
Volume:
Proceedings of the 28th International Conference on Computational Linguistics: Industry Track
Month:
December
Year:
2020
Address:
Online
Venue:
COLING
SIG:
Publisher:
International Committee on Computational Linguistics
Note:
Pages:
172–178
Language:
URL:
https://aclanthology.org/2020.coling-industry.16
DOI:
10.18653/v1/2020.coling-industry.16
Bibkey:
Cite (ACL):
Yousef El-Kurdi, Hiroshi Kanayama, Efsun Sarioglu Kayi, Vittorio Castelli, Todd Ward, and Radu Florian. 2020. Scalable Cross-lingual Treebank Synthesis for Improved Production Dependency Parsers. In Proceedings of the 28th International Conference on Computational Linguistics: Industry Track, pages 172–178, Online. International Committee on Computational Linguistics.
Cite (Informal):
Scalable Cross-lingual Treebank Synthesis for Improved Production Dependency Parsers (El-Kurdi et al., COLING 2020)
Copy Citation:
PDF:
https://aclanthology.org/2020.coling-industry.16.pdf