Re: [Accessibility] Call to Arms

[Eric S. Johansson]

 On 7/26/2010 8:14 AM, Steve Holmes wrote:
> On Sun, Jul 25, 2010 at 10:52:41PM -0400, Richard Stallman wrote:
> > Something like that might be independent enough of the recognizer
> > to be a valid project.  But there ARE free software packages for
> > speech recognition.  So people should develop it to work with them.
> > If users can also run it with NaturallySpeaking, that is ok,
> > as long as we don't suggest it.
> >
> > However, I think we should not include such things in THIS project,
> > because we need to focus energy on the goal of making those free
> > recognizers better.  For us, replacing important proprietary software
> > takes priority over advancing the capabilities of software.
> When I hear this statement, I feel a bit of a problem here.  Maybe
> it's a matter of pragmatics or something but if the current free
> solutions are inadaquite or don't work at all and a sutible
> replacement of this or a proprietary product like Naturally Speaking
> is going to be 7 to 10 years away, the people needing such a process
> would have no choice but to go with a Windows solution and abandon any
> hope of using gnu/linux and its other numerous advantages.  I like the
> idea if the support utilities were to be developed to be compatible
> with Naturally Speaking currently but with the firm understanding that
> a good replacement for Naturally Speaking be developed in parallel and
> maybe even given higher priority to shorten the development cycle.
> The context of the above statement of replacing proprietary software
> before enhancing existing software sounds a log like "all or nothing"
> to me.  Couldn't something like the LGPL be applied to this speech
> recognission effort? Develop the tools don't lock them to Naturally
> Speaking; maybe adjust the existing recognition layer to be compatible
> with what is expected by Naturally Speaking so a smooth transition
> could take place and soon be rid of depending on the proprietary
> objects.

thank you for expressing it this way. The basic idea I've had in the background 
for a few years has been something like this.

Interface side )----<lh bridge ]==[ rh bridge>--- (Recognizer

You replace the right-hand bridge and recognizer dependent on your needs. In the 
very beginning, the right-hand bridge and recognizer would be NaturallySpeaking 
plus something we build. As the years go by and a free software foundation 
recognizer comes to life, they can create their own right-hand bridge for 
testing and usage.

The other reason I recommend this approach is because the connection between the 
haves can be many thousands of miles. There's no reason to put the recognizer on 
the same machine as the interface. This model lets us support speech recognition 
on many more machines than you have recognizers[1] it also lets you put a 
recognizer on a host and speak to a virtual machine. It lets you put a 
recognizer inside of wine and speak to another machine.

[1]  to assist that a user put all of the recognition environment, macros, 
configuration etc. on every machine they use is the path to madness. It's bad 
enough keeping your shell environment consistent, a speech recognition 
environment is much more complex and correspondingly difficult to synchronize.   
another difficulty is transporting the audio stream from where you are to where 
the recognizer is.  Keeping all of your recognition/configuration information on 
a single machine reduces or eliminates all these problems and enhances the user 
experience.

Let me also point out that if I had the right tools that let me write code, I 
could've started on something this. Waiting for the free software foundation 
model, I wouldn't be writing code for at least 5 to 6 years and by that time I 
probably would've lost interest.