Re: Fwd: Voice recognition for GNOME

[Henrik Nilsen Omma <henrik ubuntu com>]

Calum Benson wrote:
> This is a new voice recognition project for GNOME:
> http://glec.umanizales.edu.co/index.php/corporate/proyectos/gervoice
> > I've took a look at it, and they seem to be using the IBM voice
> > recognition engine. BTW, the website is written in Spanish..
If it's the original ViaVoice engine, then I'm afraid it's not really 
usable as a dictation tool. Dragon Naturally Speaking is only just 
usable as that IMO, but is of course Windows-only.

I wrote a page about my experiences in using Naturally Speaking to pipe 
speech input from a Windows box into Gnome on Ubuntu via VNC a few 
months ago: https://wiki.ubuntu.com/SpeechRecognition

It sort of worked.

Even though I have little faith in the old ViaVoice engine, I think this 
new effort might still be worth while doing for several reasons. The 
worst thing about the Dragon system is it's user interface; it's more of 
a hindrance than a help. Editing the few mistakes that do occur is very 
clumsy and breaks your whole train of thought while dictating.

If this gnome project can show the way with a much more elegant user 
interface then I think there is a fair chance that Dragon might make 
it's engine available on Linux. After all, Linux can be configured for 
low latency which is very useful for music applications, and likely for 
voice recognition too. There is work being done on free engines of 
course, but I think these will need a lot of investment to catch up.

Of course IBM is still working away at its voice recognition technology 
deep in its labs. I expect the next generation will be very much better 
than what we have seen so far.

- Henrik