4 ms·
Since this is explicitly targeted at "the next billion users," do we have any sense of how well-optimized this is on non-English audio corpuses? I can't imagine
by whoisburbansky 6y ago
Since this is explicitly targeted at "the next billion users," do we have any sense of how well-optimized this is on non-English audio corpuses? I can't imagine that a model trained primarily on English/Western phonemes would perform as well on the rest of the world.
- PaulHoule 6y agoThey say they tested it on 70+ languages.
- beaconstudios 6y agowhich is less than the number of spoken languages in India alone.
- reaperducer 6y agoOr even in New York City public schools.
- core-questions 6y agoWonder if India will ever go through a forced linguistic convergence like China did
- sudhirj 6y agoUnlikely, there's too much pride in each local language. Might all converge on English over a couple of generations, though, but more for commercial reasons.
- whoisburbansky 6y agoAh you're right. I couldn't find it on the original link in this post, but the post links to https://ai.googleblog.com/2021/02/lyra-new-very-low-bitrate-codec-for.html https://ai.googleblog.com/2021/02/lyra-new-very-low-bitrate-..., which mentions the 70+ languages statistic under the "Fairness" section. Thanks!
- BlueTemplar 6y agoYeah, I wonder about those weird languages with lots of clicks... (though they are probably not part of the next billion)