Preprint
Scaling Speech Technology to 1,000+ Languages
Vineel Pratap et al., Meta AI (arXiv), 2023 · DOI 10.48550/arXiv.2305.13516
Read the originalRead on October 16, 2026
- What it is
- The Massively Multilingual Speech project from Meta. It built one recognition model covering over a thousand languages, using recordings of religious texts and self-supervised learning.
- How much weight it carries
- A preprint from the group that built the system. Coverage is not the same as quality: the paper reports error rates that vary widely by language and a training set of one text genre.
Facts taken from this source
- Meta's Massively Multilingual Speech project built a single multilingual speech recognition model for 1,107 languages.
“a single multilingual automatic speech recognition model for 1,107 languages” (Abstract)
- The MMS authors say current speech technology is limited to about one hundred languages, a small fraction of the over 7,000 languages spoken in the world.
“current speech technology is restricted to about one hundred languages which is a small fraction of the over 7,000 languages spoken around the world” (Abstract)