My understanding was that voice transcription was stuck at around 95% accuracy (one word in twenty wrong!), and wiki says only 80% http://en.wikipedia.org/wiki/

My understanding was that voice transcription was stuck at around 95% accuracy (one word in twenty wrong!), and wiki says only 80% http://en.wikipedia.org/wiki/Speech_recognition#Performance http://en.wikipedia.org/wiki/Speech_recognition#Performance Of course, you'll do far better in severely curtailed domains (eg. recognizing digits, since there's only 10). Is there other evidence for voice transcription improvement? Otherwise, this sounds a bit like "AI is just around the corner".