VoxForge
VoxForge
Is it time to grow the volunteer base for voxforge?
From what I've read about the problem, what you need is lots of people to record their voice so you can produce acoustic models.
Now if you really need people why don't you try and team up with the WikiMedia Foundation, if you can get even a small amount of wiki people to help out you'd vastly increase your incoming audio.
That is of course if you do need more people. You may just be looking for more programmers with better ideas for the software and processing side of things.
--- (Edited on 7/14/2008 10:11 pm [GMT-0500] by doctormo) ---
Hi doctormo,
>Is it time to grow the volunteer base for voxforge?
We are getting there...
>why don't you try and team up with the WikiMedia Foundation
Do you have a contact name the WikiMedia Foundation? Would it be better that you make the initial contact (i.e. do already have a relationship with them)?
>You may just be looking for more programmers with better ideas for the
>software and processing side of things.
We need both (people to submit speech and programmers). Though ideas are one thing, actually developing a solution that people will use is quite another (as evidenced by Edison's quote that invention requires 1% inspiration and 99% perspiration ... :) )
thanks,
Ken
--- (Edited on 7/15/2008 11:25 am [GMT-0400] by kmaclean) ---
> Do you have a contact name the WikiMedia Foundation? Would it be better that you make the initial contact (i.e. do already have a relationship with them)?
Unfortunatlly not, I'm an ubuntu guy myself so if it was Mark Shuttleworth it'd be no problem. But the Wikimedia foundation can't be too hard to get a relationship with.
I post the idea because it may be worth considering, I consider the idea of recorded voices (and even the audio books) to be something wiki people would be into.
--- (Edited on 7/15/2008 10:36 am [GMT-0500] by doctormo) ---
>>We need ... people to submit speech
I'm a Wikipedian. I see no point in contacting Foundation itself, but there are lot of articles that already have speech audio files, different accents. Audio itself usually in a good quality. Sometimes children voices, but that is rare. Huge vocabulary. Licenses are different, but mostly public domain/GFDL. Obviously GFDL is incompatible with GPL, but obtaining permission for audio file in most cases would be easy. Ogg format. So it is lossy compressed. I guess I can figure out a way to translate files back into wav.
And, I'm sort of willing to volunteer. But, it looks like somewhat time consuming project and real question - is it worth the efforts?
--- (Edited on 12/8/2008 3:36 pm [GMT-0600] by Visitor) ---
Hi TP,
>Ogg format. So it is lossy compressed. I guess I can figure out a way to
>translate files back into wav.
Thanks!
Please note that we prefer WAV format, since converting Ogg to WAV still means there is a loss of information as a result of the Ogg compression algorithms - see this FAQ
>And, I'm sort of willing to volunteer.
That would be great!
Note that the speech files need to be segmented into 10-20 word audio files. The process is described here: Automated Audio Segmentation Using Forced Alignment (Draft), and in the perldocs of this script: Audiobook.pm (this script still need some refinements, but it works).
>But, it looks like somewhat time consuming project and real question - is it
>worth the efforts?
If Linux is to be a worthy competitor to Windows or Mac on the desktop, then Linux needs speech. It's a matter of "table stakes" - i.e. in gambling, you need to put in a minimum amount of money into a game to even be able to play, the same applies to desktop Operating Systems.
I have seen comments that some organizations will not use Linux because it does not have speech as an accessibility option (though with virtualization, this might not be as great an issue as it once was...).
Ken
--- (Edited on 12/10/2008 2:06 pm [GMT-0500] by kmaclean) ---