When I’ve spoken out something incorrectly, I think it would be a nice additional feature if I was able to indicate that I’d read the sample in error. I’m guessing that this feature is already implicit? Are voice samples that are discarded by a volunteer deleted when the volunteer indicates that they’ve misread the sample? I’m not clear on the logistics regarding this, but it seems that those garbled samples would be useful as well, as a more useful voice recognition would be able to decipher misspoken bits of language.
