Re: Acoustic Models for HUD
On Mon, Feb 25, 2013 at 03:46:34PM -0600, Ted Gould wrote: On Mon, 2013-02-25 at 19:13 +0100, Siegfried-Angel Gevatter Pujals wrote: The English Voxforge models are currently packaged in julius-voxforge. There I did go with the nightly builds there, since in addition to the time and disk size (which IMHO is already enough of a reason), it needed HTK to build, which is not redistributable. It'd also be interested in more opinions though. Yes, thanks for doing that! We found it early on and it created some very good results. But yes, Julius does have redistribution problems. But the library being 4-clause BSD and the tools. 4-clause BSD (i.e. with the advertising clause) isn't a distribution problem in itself; from a brief glance at the copyright file I couldn't work out why it was in multiverse. Does anyone have background on that? However, 4-clause BSD is incompatible with the GPL. We could grant an exception for Canonical-owned code, but we'd then have to make sure we never used any external GPL code. It probably wouldn't be worth the hassle. -- Colin Watson [cjwat...@ubuntu.com] -- ubuntu-devel mailing list ubuntu-devel@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel
Acoustic Models for HUD
The publicity clause can be considered as a use restriction, and some people also complained about the choice of venue. Those problems have been raised upstream, but (last time I checked) didn't receive a response. http://lists.debian.org/debian-legal/2010/07/msg00024.html Am Dienstag, 26. Februar 2013 schrieb Colin Watson : On Mon, Feb 25, 2013 at 03:46:34PM -0600, Ted Gould wrote: On Mon, 2013-02-25 at 19:13 +0100, Siegfried-Angel Gevatter Pujals wrote: The English Voxforge models are currently packaged in julius-voxforge. There I did go with the nightly builds there, since in addition to the time and disk size (which IMHO is already enough of a reason), it needed HTK to build, which is not redistributable. It'd also be interested in more opinions though. Yes, thanks for doing that! We found it early on and it created some very good results. But yes, Julius does have redistribution problems. But the library being 4-clause BSD and the tools. 4-clause BSD (i.e. with the advertising clause) isn't a distribution problem in itself; from a brief glance at the copyright file I couldn't work out why it was in multiverse. Does anyone have background on that? However, 4-clause BSD is incompatible with the GPL. We could grant an exception for Canonical-owned code, but we'd then have to make sure we never used any external GPL code. It probably wouldn't be worth the hassle. -- Colin Watson [cjwat...@ubuntu.com] -- ubuntu-devel mailing list ubuntu-devel@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel -- Siegfried -- ubuntu-devel mailing list ubuntu-devel@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel
Re: Acoustic Models for HUD
Am Montag, 25. Februar 2013 schrieb Ted Gould : In the demo images we're using Julius, but we've gotten a lot of help from the Sphinx list this week to make it much better. Aha. So is their performance comparable? (Back when I tried them out Julius worked much better for me, but I didn't look into any kind of tuning -and I didn't even have much of a clue of what I was doing :p-). Here at university Kaldi has been recommended as the open-source implementation of choice, but I haven't got around to looking into it yet. Do you have any experience with it to share? Best, Siegfried -- Siegfried -- ubuntu-devel mailing list ubuntu-devel@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel
Re: Acoustic Models for HUD
On Tue, 2013-02-26 at 15:45 +0100, Siegfried-Angel Gevatter Pujals wrote: Am Montag, 25. Februar 2013 schrieb Ted Gould : In the demo images we're using Julius, but we've gotten a lot of help from the Sphinx list this week to make it much better. Aha. So is their performance comparable? (Back when I tried them out Julius worked much better for me, but I didn't look into any kind of tuning -and I didn't even have much of a clue of what I was doing :p-). We were clearly in the same state as we did get better performance with Julius. But a little help from the experts has gone a long way :-) Here at university Kaldi has been recommended as the open-source implementation of choice, but I haven't got around to looking into it yet. Do you have any experience with it to share? No, I haven't looked into it enough to have a solid opinion. Ted signature.asc Description: This is a digitally signed message part -- ubuntu-devel mailing list ubuntu-devel@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel
Re: Acoustic Models for HUD
On Tue, Feb 26, 2013 at 03:38:22PM +0100, Siegfried-Angel Gevatter Pujals wrote: The publicity clause can be considered as a use restriction, and some people also complained about the choice of venue. Those problems have been raised upstream, but (last time I checked) didn't receive a response. http://lists.debian.org/debian-legal/2010/07/msg00024.html Choice of venue is accepted as free in Debian and Ubuntu (despite misgivings expressed that this is a bad thing in a free software license). The publicity clause is strange, and does indeed go farther than the standard BSD license. I'm not sure if that's actually something that should make it non-free. Evidently, the Debian ftp team haven't ruled on this - despite being discussed on debian-legal, the package is not in the Debian archive, nor in the Debian NEW queue? -- Steve Langasek Give me a lever long enough and a Free OS Debian Developer to set it on, and I can move the world. Ubuntu Developerhttp://www.debian.org/ slanga...@ubuntu.com vor...@debian.org Am Dienstag, 26. Februar 2013 schrieb Colin Watson : On Mon, Feb 25, 2013 at 03:46:34PM -0600, Ted Gould wrote: On Mon, 2013-02-25 at 19:13 +0100, Siegfried-Angel Gevatter Pujals wrote: The English Voxforge models are currently packaged in julius-voxforge. There I did go with the nightly builds there, since in addition to the time and disk size (which IMHO is already enough of a reason), it needed HTK to build, which is not redistributable. It'd also be interested in more opinions though. Yes, thanks for doing that! We found it early on and it created some very good results. But yes, Julius does have redistribution problems. But the library being 4-clause BSD and the tools. 4-clause BSD (i.e. with the advertising clause) isn't a distribution problem in itself; from a brief glance at the copyright file I couldn't work out why it was in multiverse. Does anyone have background on that? However, 4-clause BSD is incompatible with the GPL. We could grant an exception for Canonical-owned code, but we'd then have to make sure we never used any external GPL code. It probably wouldn't be worth the hassle. -- Colin Watson [cjwat...@ubuntu.com] -- ubuntu-devel mailing list ubuntu-devel@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel signature.asc Description: Digital signature -- ubuntu-devel mailing list ubuntu-devel@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel
Re: Acoustic Models for HUD
On Tue, Feb 26, 2013 at 03:38:22PM +0100, Siegfried-Angel Gevatter Pujals wrote: The publicity clause can be considered as a use restriction, and some people also complained about the choice of venue. Those problems have been raised upstream, but (last time I checked) didn't receive a response. Mm, right. Ted's description of this as 4-clause BSD misled me a bit: the publicity clause here is indeed rather more onerous than the traditional BSD advertising clause, and I agree that this belongs in multiverse as a result. -- Colin Watson [cjwat...@ubuntu.com] -- ubuntu-devel mailing list ubuntu-devel@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel
Acoustic Models for HUD
Howdy, As some folks may have noticed we're working on a voice input feature in HUD. Part of what that requires is acoustic models to be available to understand the speech coming in. Currently in Ubuntu there are a couple of these, but we need to get to the point of providing for various languages and having a way to update these continuously as the data gets better. So that leads to the question: How do we want these to look in Ubuntu? The best open source for training data appears to be Voxforge, a collection of samples based on known text. These samples can then be used to compile the acoustical model that the various libraries need. This takes significant amounts of CPU time. Their most complete language is English, which has about 100 hours of audio, and takes about 10 CPU hours to compile the models that Sphinx needs. While English is the most complete, I think it's important to realize that the best/worst case scenario that supports all languages well could result in easily over a thousand hours of CPU time. So if we think of things in the classic source vs. binary split, it seems like the Voxforge data is the source and we should make a source package that then builds these binary models. But, at some level, we're just exchanging binary data (sound files) for different binary files (acoustic models). Would it make more sense to package something like the Voxforge nightly builds for use in Ubuntu? I'd love to hear people's thoughts on this. I'm leaning towards putting the Voxforge data as a source package, as it is our source, but I'm worried about the impact it may have on rebuilding the archive. Thanks, Ted signature.asc Description: This is a digitally signed message part -- ubuntu-devel mailing list ubuntu-devel@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel
Acoustic Models for HUD
Hi Ted, It's great to hear that voice recognition in Ubuntu is finally getting some love :). The English Voxforge models are currently packaged in julius-voxforge. There I did go with the nightly builds there, since in addition to the time and disk size (which IMHO is already enough of a reason), it needed HTK to build, which is not redistributable. It'd also be interested in more opinions though. Out of curiosity, what's the plan for voice recognition in Ubuntu? Sphinx/Julius/Kaldi? Regards, Siegfried Am Montag, 25. Februar 2013 schrieb Ted Gould : ** Howdy, As some folks may have noticed we're working on a voice input feature in HUD. Part of what that requires is acoustic models to be available to understand the speech coming in. Currently in Ubuntu there are a couple of these, but we need to get to the point of providing for various languages and having a way to update these continuously as the data gets better. So that leads to the question: How do we want these to look in Ubuntu? The best open source for training data appears to be Voxforgehttp://www.voxforge.org, a collection of samples based on known text. These samples can then be used to compile the acoustical model that the various libraries need. This takes significant amounts of CPU time. Their most complete language is English, which has about 100 hours of audio, and takes about 10 CPU hours to compile the models that Sphinx needs. While English is the most complete, I think it's important to realize that the best/worst case scenario that supports all languages well could result in easily over a thousand hours of CPU time. So if we think of things in the classic source vs. binary split, it seems like the Voxforge data is the source and we should make a source package that then builds these binary models. But, at some level, we're just exchanging binary data (sound files) for different binary files (acoustic models). Would it make more sense to package something like the Voxforge nightly buildshttp://www.repository.voxforge1.org/downloads/Nightly_Builds/for use in Ubuntu? I'd love to hear people's thoughts on this. I'm leaning towards putting the Voxforge data as a source package, as it is our source, but I'm worried about the impact it may have on rebuilding the archive. Thanks, Ted -- Siegfried -- ubuntu-devel mailing list ubuntu-devel@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel
Re: Acoustic Models for HUD
On Mon, Feb 25, 2013 at 07:13:43PM +0100, Siegfried-Angel Gevatter Pujals wrote: It's great to hear that voice recognition in Ubuntu is finally getting some love :). The English Voxforge models are currently packaged in julius-voxforge. There I did go with the nightly builds there, since in addition to the time and disk size (which IMHO is already enough of a reason), it needed HTK to build, which is not redistributable. It'd also be interested in more opinions though. In terms of freeness of the OS, depending on non-redistributable tools for building the data files is more of an issue than whether we actually process them at package build time. Is this a julius-specific requirement, or does it also affect sphinx? -- Steve Langasek Give me a lever long enough and a Free OS Debian Developer to set it on, and I can move the world. Ubuntu Developerhttp://www.debian.org/ slanga...@ubuntu.com vor...@debian.org Am Montag, 25. Februar 2013 schrieb Ted Gould : ** Howdy, As some folks may have noticed we're working on a voice input feature in HUD. Part of what that requires is acoustic models to be available to understand the speech coming in. Currently in Ubuntu there are a couple of these, but we need to get to the point of providing for various languages and having a way to update these continuously as the data gets better. So that leads to the question: How do we want these to look in Ubuntu? The best open source for training data appears to be Voxforgehttp://www.voxforge.org, a collection of samples based on known text. These samples can then be used to compile the acoustical model that the various libraries need. This takes significant amounts of CPU time. Their most complete language is English, which has about 100 hours of audio, and takes about 10 CPU hours to compile the models that Sphinx needs. While English is the most complete, I think it's important to realize that the best/worst case scenario that supports all languages well could result in easily over a thousand hours of CPU time. So if we think of things in the classic source vs. binary split, it seems like the Voxforge data is the source and we should make a source package that then builds these binary models. But, at some level, we're just exchanging binary data (sound files) for different binary files (acoustic models). Would it make more sense to package something like the Voxforge nightly buildshttp://www.repository.voxforge1.org/downloads/Nightly_Builds/for use in Ubuntu? I'd love to hear people's thoughts on this. I'm leaning towards putting the Voxforge data as a source package, as it is our source, but I'm worried about the impact it may have on rebuilding the archive. signature.asc Description: Digital signature -- ubuntu-devel mailing list ubuntu-devel@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel
Re: Acoustic Models for HUD
On Mon, 2013-02-25 at 11:41 -0800, Steve Langasek wrote: On Mon, Feb 25, 2013 at 07:13:43PM +0100, Siegfried-Angel Gevatter Pujals wrote: It's great to hear that voice recognition in Ubuntu is finally getting some love :). The English Voxforge models are currently packaged in julius-voxforge. There I did go with the nightly builds there, since in addition to the time and disk size (which IMHO is already enough of a reason), it needed HTK to build, which is not redistributable. It'd also be interested in more opinions though. In terms of freeness of the OS, depending on non-redistributable tools for building the data files is more of an issue than whether we actually process them at package build time. Is this a julius-specific requirement, or does it also affect sphinx? Julius uses the HTK models. To build the Sphinx ones we need sphinx-train (which isn't packaged yet, but we've got some test packages) and some scripts that Pete wrote. https://launchpad.net/voxforge-sphinx-train/ We haven't gone through a license audit on those yet, but I don't expect any issues. That is an interesting question though, is there a binary blob argument if we built the models offline and not in the source packages? It would seem like as long as it was documented and used free tools there wouldn't be an issue. Ted signature.asc Description: This is a digitally signed message part -- ubuntu-devel mailing list ubuntu-devel@lists.ubuntu.com Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel