General Discussion

Flat
How Can I Help?
User: EricHedekar
Date: 5/9/2014 2:13 pm
Views: 4297
Rating: 6

Hi, I'm very interested in helping this project in a long-term fashion.  I have a background in both audio recording and open source software, and  I am studying computational linguistics at school right now.  I am quite excited to help this project grow but would like to know the most productive method for my involvement.


Should I spend time recording speech?  Are AudioBooks being included in the corpus?  Can Creative Commons books be included or are the licences incompatible?


Should I spend time vetting recordings for transcription errors?

Should I spend time writing processing scripts?

Should I be poking all of my friends who speak other languages (French, German, Arabic, etc..) to contribute?

Where is this project at and what are the major boulders that need an extra hand to push up the hill?

--- (Edited on 5/9/2014 2:13 pm [GMT-0500] by EricHedekar) ---

Re: How Can I Help?
User: nsh
Date: 5/12/2014 2:25 pm
Views: 345
Rating: 6

Dear Eric

The most valuable help would be to write some code :) but datacollection could also help.

> Should I spend time recording speech? 

No

> Are AudioBooks being included in the corpus? 

Yes

> Can Creative Commons books be included or are the licences incompatible?

Even if they are not compatible they are very useful if you find some amount of them and help to process.
Actually there are quite many US English audiobooks around already, the question is how to process them. We are working on it but not there yet.

> Should I spend time vetting recordings for transcription errors?

No, it's better to implement this in the code

> Should I spend time writing processing scripts?

Yes

> Should I be poking all of my friends who speak other languages (French, German, Arabic, etc..) to contribute?

Sure

> Where is this project at and what are the major boulders that need an extra hand to push up the hill?

There are many tasks as usual, some of them are extremely interesting like distributed grid-like model training. If you are willing to contribute I'm sure you'll find many opportunities.
Add me in skype if you want to discuss this or join #cmusphinx irc forum on freenode.

--- (Edited on 5/12/2014 23:25 [GMT+0400] by nsh) ---

Re: How Can I Help?
User: camdixon
Date: 12/7/2014 1:39 am
Views: 80
Rating: 3

My opinion, scripts for sure.  I was working on making a script to add into a separate page.  The script did extract and play the recordings for users on this webpage (after some screening scripts ran) which would REALLY help voxforge as it would allow for distributed and user accepted processing.  Basically, you need a username / pass setup at the moment to login and process the voice samples sumbitted via the site I belive.  Let me know if you're interested in adding to / working on the script.

Seems like available sources are not the problem it's processing them all.  

Not sure which languages need more voice files or which are priority.  The more the merrier?

--- (Edited on 12/7/2014 1:39 am [GMT-0600] by camdixon) ---

--- (Edited on 12/7/2014 1:41 am [GMT-0600] by camdixon) ---

Re: How Can I Help?
User: camdixon
Date: 12/7/2014 1:46 am
Views: 1551
Rating: 3

Erik,

You should also see a recent post on this forum here

http://www.voxforge.org/home/forums/message-boards/audio-discussions/why-is-there-so-much-speech-sitting-in-the-waiting-list#ZfuMlZ8N17bSi2LNKtGMxQ

 

Where ken (one of the organizers I believe) details why so many speech samples are awaiting processing.  He says they crash the acoustic models.

">buuuut, the submission process is still marking loads of entries as

>"awaiting processing" for some reason?

I have not got around to processing them - we get submissions that will crash the acoustic model creation scripts for a variety of reasons.  The verification process is only semi-automated, and very tedious..." 

--- (Edited on 12/7/2014 1:46 am [GMT-0600] by camdixon) ---

PreviousNext