Re:High-performance speech recognition with no supervision a...
by shakascchen on Saturday May 22, 2021 @01:25AM
(#27243757)
It seems that the discriminator tests if the output of the generator obeys the pattern of a language. Isn't it still possible that after training the generator outputs the text wrong for the input? On the other words, the output is understandable but actually not what the sound indicate?
[ Parent
| on HN ]
Re:High-performance speech recognition with no supervision a...
by unityByFreedom on Saturday May 22, 2021 @01:52AM
(#27243903)
Context may help determine the intended word. Certainly any such system can be wrong sometimes.