>Since the solution I am using is server based and there is no distribution,

>does that mean I am exempt as the same as GPL software running on a

>server is?

My understanding is that the GPL only applies to the 'distribution' of software and does not prevent the 'use' of software.

If you are using VoxForge audio to create an acoustic model for  server-based speech recognition, then, as long as you are not distributing the acoustic model to your users, you should be not be in breach of the GPL. 

Ken

P.S. I am not a lawyer and this is not legal advice

--- (Edited on 1/25/2010 3:46 pm [GMT-0500] by kmaclean) ---

I agree, Ken. My understanding is as follows (of course, I might be wrong):

source code = wav files + corresponding prompts file

object code = compiled speech model (hmmdefs, model.dict, ...)

If you distribute the object code, you probably have to give access to the source code. But if you use the object code only on your server, without making the object code available for download, you should be OK.

This is my understanding of the GPL (of course, I can't give you legal advice).

***

The interesting question would be the following: Are you allowed to distribute only the object code (without the corresponding wav files) under the GPL? Probably not:

"Convey the object code by offering access from a designated place (gratis or for a charge), and offer equivalent access to the Corresponding Source in the same way through the same place at no further charge."

As long as you don't convey the object code, you should be fine.

I would say: just do it, ecret. It is a difficult question, but I think that as long as you don't distribute the object code itself, there is absolutely no legal problem (this is just my opinion).

***

By the way: I am planning to distribute compiled speech models (that could be seen as object code) under the GPL. Even object code could be used as source code (for model adaption). At the moment, I don't have a legal problem because I am only using my own voice for the creation of acoustic models. But in the future, we may have to discuss this difficult question.

The GPL should not make our goal - distribution of free speech models - unusable. So even a compiled speech model (= object code) can serve as source code. If it is the intention to provide a compiled speech model as source code, this should not be seen as a violation of the GPL (my opinion).

It may be necessary that we continue this GPL discussion during the next couple of months. At the moment, I don't have a problem. But I am planning to use sam for the creation and adaption of acoustic models. And of course it is my intent to publish the compiled speech models as source (!) code so that other users can use it.

So what is with object code that can be used as source code?

--- (Edited on 2010-01-26 2:38 am [GMT-0600] by ralfherzog) ---

i Ralf,

>I am planning to distribute compiled speech models (that could be seen as

>object code) under the GPL.

What do you mean by speech model... language model, acoustic model, or both?

> don't have a legal problem because I am only using my own voice for the

>creation of acoustic models.

The GPL gets its force from Copyright law.  If you own the Copyright in a work, then you can pretty much do what you want with it, including licencing it under the GPL.

>So even a compiled speech model (= object code) can serve as source

>code.

From the GPL FAQ:

Can I use the GPL for something other than software?

You can apply the GPL to any kind of work, as long as it is clear what constitutes the “source code” for the work. The GPL defines this as the preferred form of the work for making changes in it.

Ken

--- (Edited on 1/26/2010 10:32 pm [GMT-0500] by kmaclean) ---

Hi Ken,

"What do you mean by speech model... language model, acoustic model, or both?"

I don't distinguish any more between language model and acoustic model. My goal is to offer my own "speech model" that the user can e.g. use with simon.

Recently, I tried a static model from Voxforge. It was possible to use HTK_AcousticModel-2010-02-03_16kHz_16bit_MFCC_O_D.tgz in conjunction with simon (the current development version of simon allows to set paths to hmmdefs / tiedlist / macros / stats).

So obviously, I used an acoustic model from Voxforge (and not a language model).

By the way, if someone is interested in the current development version of simon, you can type into the Ubuntu terminal: $ git clone git://speech2text.git.sourceforge.net/gitroot/speech2text/speech2text

Then you can try the acoustic model from Voxforge HTK_AcousticModel-2010-02-03_16kHz_16bit_MFCC_O_D.tgz with simon. It should work.

Sorry for being a little bit off-topic, but I find it important that it is possible to use the Voxforge acoustic model with simon (development version).

The file HTK_AcousticModel-2010-02-03_16kHz_16bit_MFCC_O_D.tgz serves as source code. So in my opinion, I would be permitted to distribute exactly this .tgz-file without explicitly referencing to the prompts and the original wav files.

Greetings,

Ralf

--- (Edited on 2010-02-06 4:51 am [GMT-0600] by ralfherzog) ---

Hi Ralf,

>The file HTK_AcousticModel-2010-02-03_16kHz_16bit_MFCC_O_D.tgz

>serves as source code. So in my opinion, I would be permitted to

>distribute exactly this .tgz-file without explicitly referencing to the prompts

>and the original wav files.

But you don't own the Copyright to the original wav or prompt files.  So I am not sure I understand how you can legally declare a derivative work (like the VoxForge acoustic model) to be "source code", when your have not created the work yourself.

I understand that you are using it as input to a process to create a static acoustic model, and therefore it might be  considered as 'source' from a programming perspective, but the Copyright owners of all the speech that was used to create the VoxForge acoustic model that you are using as 'source' for adapting and creating your own 'static' models might have something to say about it.

It seems like you are in essence trying to defeat the whole purpose of using the GPL license on the VoxForge audio and prompts, which I don't think is your intention.

All you need to do is provide a link to point to the VoxForge site for the source audio (or some other repository which contains the source audio) when you distribute a VoxForge acoustic model.  The FSF FAQ says:

Can I put the binaries on my Internet server and put the source on a different Internet site?

The GPL says you must offer access to copy the source code "from the same place"; that is, next to the binaries. However, if you make arrangements with another site to keep the necessary source code available, and put a link or cross-reference to the source code next to the binaries, we think that qualifies as "from the same place".

...

>... but I find it important that it is possible to use the Voxforge acoustic

>model with simon (development version).

I agree 100%.  It is very important that the VoxForge acoustic models can be used with Simon, or any other dialog manager that makes things easier for users.  I think all that you need is a link to the source audio and prompts, and in my view, everything will be kosher from a GPL perspective.

Ken

IANAL & TINLA

--- (Edited on 2/6/2010 3:53 pm [GMT-0500] by kmaclean) ---