VEDICS Speech Assistant

[Eric S. Johansson]

On 5/22/2010 3:03 AM, Kenny Hitt wrote:
> Hi.
>
> On Fri, May 21, 2010 at 04:21:53PM -0400, Eric S. Johansson wrote:
>> On 5/21/2010 11:04 AM, Nischal Rao wrote:
> <snip>
>
>>>
>>> Currently the software doesn't support the dictation facility. However,
>>> we are planning to add this feature in the future.
>>> The best part of this software is that it is speaker independent, no
>>> training is required and it can recognize words not present in the
>>> English dictionary.
>>>
>>> Currently it works well on ubuntu 9.10 and ubuntu 10.04
>>>
>>> You can find the source code at : http://sourceforge.net/projects/vedics/
>>
>> very nice.  have you thrown away your keyboard yet?  please do so and send a
>> message to the list without keyboard.
>>
> Before you post such a negative message, you should really read first.
> This is not even a stable tarball release yet.  The author stated clearly
> dictation wasn't available, but is planned to be added.
> If he had claimed that you could do dictation, your post would make since, but since
> he didn't, you look like a winy ass.
> When a project like this is still at such an early stage, bad attitude will cause a developer
> to wonder if the trouble is really worth it.

those who are unaware of history are doomed to repeat it... badly

This is about the 5th time I've seen this sort of project get started. I've seen 
every single commercial equivalent fail. I've watched people get excited over 
and over again thinking that at IVR level recognition engine can be used to 
replicate NaturallySpeaking functionality only to have their hopes crushed and 
energy wasted when they discover the two engine types are radically different.

This is not to say the project could be useful in a particular problem domain 
such as robotic control or command and control by telephone it's just that 
history shows that this idea has failed when applied to accessibility needs 
because the vast majority of speech recognition use by disabled person is the 
creation of text, not noodling around on the desktop. After all, what value does 
setting the font have when your hands won't let you type the text.

I've been involved in the Linux desktop recognition issue for a very long time. 
  I've had conversations with senior management at Dragon Systems (pre-buyout) 
on the market strategy (they still can't figure out how to make money in Linux 
today because it will only cannibalize the Windows market and depressed 
pricing). I've participated in the creation of a nonprofit focused on creating 
Linux desktop speech recognition systems and watched its dissolution because we 
couldn't get the technology and, we couldn't get sufficient technical support 
from the OSS community to build what was needed. They wanted to build something 
based on sphinx or Julius, both of which would not meet our needs. this opinion 
of suitability came from the developers of the SR engine projects.

If by being a whiny ass, you mean being a historian and making people aware of 
how they are wasting their time, unfairly raising people's hopes and building 
something for which they've not even studied the basic use case, then yes, I'll 
be a whiny ass.

Ever since I've been injured, I've been watching upper extremity disabled people 
jumping up and down, waving their hands or something that doesn't hurt as badly, 
saying "hey hey hey, we need help over here!"  unfortunately, the people who can 
write code are somewhere else saying "this should be useful because we know how 
to write this code." End result being being we get nothing we can use and 
developers think we are a bunch of ungrateful shits because we don't think their 
projects are wonderful.

A further bit of insanity comes when someone asks for help integrating an 
open-source program (Emacs via VR-mode) to work with NaturallySpeaking. If we 
get a response, it's frequently "we can't do that, it would encourage people to 
use proprietary software". Head-desk. There seems to be a blind spot recognizing 
that what it takes to make a good speech user interface is complex, in many ways 
far more complex than almost all accessibility interfaces put together. Speech 
UIs really any need to be built at first and let the recognition engine come 
second or third in your priority list because the first priority should be 
making an environment work for speech recognition users.

a related winey ass bit is that I've got ideas for speech UIs, I can't implement 
them because my hands are broken. I need someone to be a coding Buddy to work 
with as a team (two minds, one set of hands) to bring my ideas to fruition and 
find them solve the problems with them. This pair problem is also one of the 
reasons why speech recognition users have made so little progress improving 
their own lot over the past decade. We can't find developers with hands who are 
willing to truly listen to us. As result, we pick out code slowly and with lots 
of errors and sometimes, it just gets to be too much, too exhausting and 
projects fall by the wayside.

hopefully this is my last whiny ass bit. OSS is nice, OSS should be added 
incrementally but if the ideology gets in the way of people being able to make 
money, to live independently then it should be sidelined until it no longer gets 
in the way.  Otherwise, why should you bother with accessibility at all?


> One note to those people who have recently started to get a ubuntu accessibility group going again:
> you really need to subscribe to gnome-accessibility.  All the developments in accessibility are happening
> in upstream gnome and not ubuntu.
> There was a more complete discussion about this particular app on gnome-accessibility.

do you have a URL to the archives or a rough of conversation time so I can take 
a look?