YouTube-ASL: A Large-Scale, Open-Domain American Sign Language-English Parallel Corpus

Abstract

Machine learning for sign languages is bottlenecked by data. In this paper, we present YouTube-ASL, a large-scale, open-domain corpus of American Sign Language (ASL) videos and accompanying English captions drawn from YouTube. With ~1000 hours of videos and >2500 unique signers, YouTube-ASL is ~3x as large and has ~10x as many unique signers as the largest prior ASL dataset. We train baseline models for ASL to English translation on YouTube-ASL and evaluate them on How2Sign, where we achieve a new fine-tuned state of the art of 12.397 BLEU and, for the first time, nontrivial zero-shot results.

PDF NeurIPS OpenReview Semantic Scholar

Cite

Text

Uthus et al. "YouTube-ASL: A Large-Scale, Open-Domain American Sign Language-English Parallel Corpus." Neural Information Processing Systems, 2023.

Markdown

[Uthus et al. "YouTube-ASL: A Large-Scale, Open-Domain American Sign Language-English Parallel Corpus." Neural Information Processing Systems, 2023.](https://mlanthology.org/neurips/2023/uthus2023neurips-youtubeasl/)

BibTeX

@inproceedings{uthus2023neurips-youtubeasl,
  title     = {{YouTube-ASL: A Large-Scale, Open-Domain American Sign Language-English Parallel Corpus}},
  author    = {Uthus, Dave and Tanzer, Garrett and Georg, Manfred},
  booktitle = {Neural Information Processing Systems},
  year      = {2023},
  url       = {https://mlanthology.org/neurips/2023/uthus2023neurips-youtubeasl/}
}