Click here to register.

English Speech Files

Flat
Comma Gets a Cure
User: pmahoney
Date: 10/10/2006 7:47 pm
Views: 3008
Rating: 25

Hey all,

let me know how this turned out. This could be one of the most important projects in free software.

Pat 

--- (Edited on 10/10/2006 7:47 pm [GMT-0500] by pmahoney) ---

pmahoney-10102006.tar.gz pmahoney-10102006.tar.gz

Notice: many prompts in "English Speech Files" were adapted from the prompt files contained in the CMU_ARCTIC speech synthesis database, which were in turn derived from out-of-copyright texts from Project Gutenberg, by the FestVox project at the Language Technologies Institute at Carnegie Mellon University.

Re: Comma Gets a Cure
User: kmaclean
Date: 10/10/2006 10:27 pm
Views: 245
Rating: 22

thanks for the submission.

It needed to be approved by the moderator before it would display.

Ken 

--- (Edited on 10/10/2006 11:27 pm [GMT-0400] by kmaclean) ---


Notice: many prompts in "English Speech Files" were adapted from the prompt files contained in the CMU_ARCTIC speech synthesis database, which were in turn derived from out-of-copyright texts from Project Gutenberg, by the FestVox project at the Language Technologies Institute at Carnegie Mellon University.

Re: Comma Gets a Cure
User: kmaclean
Date: 10/12/2006 11:45 am
Views: 242
Rating: 21

Please submit as much speech as you can!

The target amount of audio for release 1.0 of the VoxForge Speech Audio Repository is 140 hours of speech (why? this is the same amount of audio used by the Sphinx Acoustic Models used by the Sphinx Group Speech Recognition Engines.)

thanks,

Ken 

--- (Edited on 10/12/2006 12:45 pm [GMT-0400] by kmaclean) ---


Notice: many prompts in "English Speech Files" were adapted from the prompt files contained in the CMU_ARCTIC speech synthesis database, which were in turn derived from out-of-copyright texts from Project Gutenberg, by the FestVox project at the Language Technologies Institute at Carnegie Mellon University.

Re: Comma Gets a Cure
User: pmahoney
Date: 10/12/2006 3:27 pm
Views: 264
Rating: 23

I think we could easily pull that off with a little time...a large community could also speed it along. 140 people at 1 hour-->!!  if we don't get that many participants, more prompts might be necessary.

Would it be better to have small amounts of speech from many people, or a lot of speech from a few people?

Pat 

--- (Edited on 10/12/2006 3:27 pm [GMT-0500] by pmahoney) ---


Notice: many prompts in "English Speech Files" were adapted from the prompt files contained in the CMU_ARCTIC speech synthesis database, which were in turn derived from out-of-copyright texts from Project Gutenberg, by the FestVox project at the Language Technologies Institute at Carnegie Mellon University.

Re: Comma Gets a Cure
User: kmaclean
Date: 10/18/2006 10:38 am
Views: 503
Rating: 27

Sorry for the late reply, I was relying on receiving email notices for new posts, but I must have missed yours.

Pat said: "Would it be better to have small amounts of speech from many people, or a lot of speech from a few people?"

There are two aspects to this question: technical and social. 

From a technical perspective, you need both.  You need good coverage of all the different phones and triphones in the english language, so having broad coverage from a few people would permit this.  But to make make the Acoustic Model "Speaker Independent", you need smaller contributions from many other people.

From a social perspective, we need to be realistic about what people are willing to contribute, and likely small amounts of speech from many people are all we can reasonably expect.  This is our reality. 

However, to get around this have ensure we have a broad coverage of all the different phones and triphones in the english language, I am looking at supplementing user contributions of audio with contributions made by users who create audio books (e.g. Librivox,  Guttenberg Project, Spoken Wikipedia, etc.).  This will take a bit of time to set up, but it is in the works,

Ken 

 

--- (Edited on 10/18/2006 11:38 am [GMT-0400] by kmaclean) ---


Notice: many prompts in "English Speech Files" were adapted from the prompt files contained in the CMU_ARCTIC speech synthesis database, which were in turn derived from out-of-copyright texts from Project Gutenberg, by the FestVox project at the Language Technologies Institute at Carnegie Mellon University.

PreviousNext