Re: Request for Assistance Finding FLOSS Speech Recognition Software

[Peter Grasch <grasch simon-listens org>]

Am Dienstag, 26. Juni 2012, 17:07:24 schrieb Jason White:
> Mario Fux <a11y-ml unormal org> wrote:
> > This summer (end of September) we plan to do an Accessibility meeting in
> > Randa/Switzerland with people from both Gnome and KDE.
> 
> Excellent. Can the recognizer be trained or customized to meet the needs of
> people whose speech may be affected by a disability, and which therefore
> falls outside the range of what the tool is trained by default to process?
> 
> I don't know anything about the inner workings of speech recognition
> systems, other than that they're based on statistical models and machine
> learning algorithms. A free and open-source recognizer could be used by
> anybody as the basis of a customization project, of course, whereas a
> proprietary system probably couldn't (at least not in an open academic or
> research setting) since the code is proprietary and likely to be available,
> if at all, only under very restrictive non-disclosure terms.

Hi,

Simon dev here.

Simon actually doesn't ship with a default model but let's the user build his 
own model from the ground up instead. This was done especially for people with 
dysarthria or otherwise unusual speech patterns.

However, because you train Simon from the ground up, the vocabulary 
constraints are tight if you don't want to spend *a lot* of time on training.
That means that while command & control can work very well with a relatively 
short period of training, dictation is - for now at least - not possible. 
Still, we do provide a voice operated keyboard which can be used to spell out 
words and have once also set up a voice controlled Dasher that also worked 
surprisingly well.

Simon is meant for tech savvy end users so you can try to set it up for 
yourself (Support forum: http://forum.kde.org/viewforum.php?f=216). The 
released version also ships with an extensive manual.

Let me know if I can help.

Best regards,
Peter