Re:Accents in Latent Spaces: How AI Hears Accent Strength in...
by Nopoint2 on Sunday May 11, 2025 @12:31AM
(#43951391)
The gibberish it produces shows that something quite different is being encoded.
It completely fails with Spanish. There is no reason why it would, if it used phonemes.
They think that the footnote explains the poor accuracy, but that is in fact how phonemes supposedly work.
I used to think the same until I came across the VQ-VAE paper [1], where the goal was to let the model learn the smallest units of speech without any supervision. Interestingly, the model ended up learning units that closely aligned with the phonemes we recognize today.
[1] - https://arxiv.org/pdf/1711.00937