IWLST 23 Code

Winning submission from low-resource ST track

About

NAVER LABS Europe submitted a single multilingual speech translation system that achieved the top performance in the IWSLT 2023 low-resource track for the Tamasheq–French and Quechua–Spanish language pairs. Our model leveraged pre-trained speech and text models, and it was implemented on top of the pasero toolkit.

Link to resources

  • Training recipe: GitHub

Citing us

For citing us, please use the bibtex below:

@inproceedings{gow-smith-etal-2023-naver,
    title = "{NAVER} {LABS} {E}urope{'}s Multilingual Speech Translation Systems for the {IWSLT} 2023 Low-Resource Track",
    author = "Gow-Smith, Edward  and
      Berard, Alexandre  and
      Boito, Marcely Zanon  and
      Calapodescu, Ioan",
    editor = "Salesky, Elizabeth  and
      Federico, Marcello  and
      Carpuat, Marine",
    booktitle = "Proceedings of the 20th International Conference on Spoken Language Translation (IWSLT)",
    month = jul,
    year = "2023",
    address = "Toronto, Canada (in-person and online)",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.iwslt-1.10.pdf",
    doi = "10.18653/v1/2023.iwslt-1.10",
    pages = "144--158"
}