Terry Blanton wrote:
> >Just curious: did you type this reply or did your voice recog s/w
> >make up 'ememies'?
>
> I managed to type that. Voice input seldom makes that kind of error.
I wondered if the s/w used a dictionary or simply phonemes (not simple)?
I do not know how it works, but it must use a dictionary at some
stage in the analysis, because it never comes up with nonexistent
words. When I say "emeny" it pauses for a while and then transcribes
that as "in many." When I say words in Japanese it comes up with all
kinds of implausible but correctly spelled English gibberish. (I can
train it for Japanese words such as "Iwamura.")
(Looking toward a multi-lingual s/w package.)
I do not think a person can use multilingual voice input unless he
speaks the second language indistinguishably like a native. I suppose
a foreign accent would discombobulate the software. But I do not know
that for a fact. I have never tried the Japanese version of NaturallySpeaking.
- Jed