Principal investigator (ÚFAL): 

Dušan Variš

Project Manager (ÚFAL): 

Hana Kubištová

Provider: 

GAUK

Grant id: 

1140218

Duration: 

2018-2020

Tags: 

Machine Translation

People: 

Ondřej Bojar

The goal of the project is to improve existing methods for in-depth error analysis of the current state-of-the-art machine translation (MT) systems. The projects consists of three parts: In the first part, we focused on improving the current best approaches towards MT output evaluation studying mainly knowledge of morphology or syntax. In the second part, we explored options for using continuous representation of words (a.k.a. embeddings) for evaluation of the MT output. In the third part, we focused on MT evaluation. With regard for a long-term research, we focused on the literary domain, which is still difficult for the current MT systems. The project focused on reseach using English-Czech language pair. Due to the increasing interest in the machine translation between English and Asian languages, the we also tried to perform contrastive experiments on the English-Japanese language pair.

Publications

• Dušan Variš, Satoshi Nakamura, Katsuhito Sudoh (2020): Image Captioning with Visual Object Representations Groundedin the Textual Modality (Electronic). In: ArXiv.org Computing Research Repository, ISSN 2331-8422 (url, local PDF)
• Dušan Variš, Ondřej Bojar (2019): Unsupervised Pretraining for Neural Machine Translation Using Elastic Weight Consolidation. In: Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics: Student Research Workshop, pp. 130-135, Association for Computational Linguistics, Stroudsburg, PA, USA, ISBN 978-1-950737-47-5 (pdf, bibtex)
• Dušan Variš, Natalia Klyueva (2018): Improving a Neural-based Tagger for Multiword Expression Identification. In: Proceedings of the 11th International Conference on Language Resources and Evaluation (LREC 2018), pp. 2526-2532, European Language Resources Association, Miyazaki, Japan, ISBN 979-10-95546-00-9 (pdf, bibtex)
• Jindřich Helcl, Jindřich Libovický, Dušan Variš (2018): CUNI System for the WMT18 Multimodal Translation Task. In: Proceedings of the Third Conference on Machine Translation, Volume 2: Shared Tasks, pp. 622-629, Association for Computational Linguistics, Stroudsburg, PA, USA, ISBN 978-1-948087-81-0 (pdf, bibtex)
• Tom Kocmi, Dušan Variš, Ondřej Bojar (2018): CUNI Basque-to-English Submission in IWSLT18. In: Proceedings of the International Workshop on Spoken Language Translation, pp. 142-146, Karlsruhe Institute of Technology, Karlsruhe, Germany (pdf, bibtex)