NVIDIA NeMo Offline Speech Translation Systems for IWSLT 2023

Oleksii Hrinchuk, Vladimir Bataev, Evelina Bakhturina, Boris Ginsburg


Abstract
This paper provides an overview of NVIDIA NeMo’s speech translation systems for the IWSLT 2023 Offline Speech Translation Task. This year, we focused on end-to-end system which capitalizes on pre-trained models and synthetic data to mitigate the problem of direct speech translation data scarcity. When trained on IWSLT 2022 constrained data, our best En->De end-to-end model achieves the average score of 31 BLEU on 7 test sets from IWSLT 2010-2020 which improves over our last year cascade (28.4) and end-to-end (25.7) submissions. When trained on IWSLT 2023 constrained data, the average score drops to 29.5 BLEU.
Anthology ID:
2023.iwslt-1.42
Volume:
Proceedings of the 20th International Conference on Spoken Language Translation (IWSLT 2023)
Month:
July
Year:
2023
Address:
Toronto, Canada (in-person and online)
Venue:
IWSLT
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
442–448
Language:
URL:
https://aclanthology.org/2023.iwslt-1.42
DOI:
Bibkey:
Cite (ACL):
Oleksii Hrinchuk, Vladimir Bataev, Evelina Bakhturina, and Boris Ginsburg. 2023. NVIDIA NeMo Offline Speech Translation Systems for IWSLT 2023. In Proceedings of the 20th International Conference on Spoken Language Translation (IWSLT 2023), pages 442–448, Toronto, Canada (in-person and online). Association for Computational Linguistics.
Cite (Informal):
NVIDIA NeMo Offline Speech Translation Systems for IWSLT 2023 (Hrinchuk et al., IWSLT 2023)
Copy Citation:
PDF:
https://preview.aclanthology.org/nodalida-main-page/2023.iwslt-1.42.pdf