Improving voice dictation results?
Recently bought a MacBook Pro (Retina) and working (or trying to work) with the voice dictation. However the quality of the results actually seems to be getting worse over time, whereas it's supposed to get better. I suspect that I am giving it confusing feedback, but can anyone explain how it is working or how it is supposed to work?
My original theory was that sometimes I am editing silently but with the dictation active, and it was interpreting that as incorrect feedback about recognition errors that weren't there. To prevent this, I tried making sure that the voice input was disengaged before making any editorial changes, but that did not seem to help.
The feedback coming from the Mac's side is also somewhat confusing, but it definitely seems as though the Mac is refining its interpretation for several seconds. After I have said a few more words, it often goes back and changes part of the text that was already displayed, so I think it is building better recognition results based on the additional context. Sometimes it winds up with ambiguous text in blue and the option to select the proper text, but it definitely seems that it offers that option less frequently these days, even though the quality of the transcription in that passage may be poor. (There are times when it seems to be working better, but I also can't find any pattern or speaking style for consistently better results. The average transcription quality seems to be near the breakeven point... Sometimes it is definitely faster than typing, but other times the corrections take longer than just typing it in the first place.)
Related topic, but sometimes I want to go back and change part of what I dictated by inserting new words in the middle, still using the dictation. However, right now I think that is a bad idea, and I am trying to avoid doing it. The Mac ought to be smart enough to figure out what is going on, and even to recognize the nearby content as relevant, but I guess not.
As a constructive suggestion for recognizing the words properly, it would be quite helpful if I could tell the Mac to play the sound it thinks it heard around some word so that I could correct the displayed words in accord with what I actually said. Perhaps with an explicit switch to a no-dictation editing mode for final polishing?
MacBook Pro with Retina display