Re:No Thanks, Google. I'll Speak for Myself
by romwell on Wednesday May 09, 2018 @07:44PM
(#17035288)
Oh gee thanks. I have an accent, ring me up when they make a bot that understands me when I don't speak in an unnatural, over-enunciated style.
Voice recognition has been improving greatly, but it is still painfully close to the Scottish Elevator[1] at times in my experience.
Duplex largely uses hems and haws like humans do - to signal that they're thinking about what to say next.
The system also sounds more natural thanks to the incorporation of speech disfluencies (e.g. “hmm”s and “uh”s). These are added when combining widely differing sound units in the concatenative TTS or adding synthetic waits, which allows the system to signal in a natural way that it is still processing. (This is what people often do when they are gathering their thoughts.) In user studies, we found that conversations using these disfluencies sound more familiar and natural.
https://ai.googleblog.com/2018/05/duplex-ai-system-for-natur...
Natural-sounding TTS improves the accuracy of speech recognition, because people tend to speak in an unnatural and over-enunciated style when they know they're talking to a bot.