I am interested in knowing if the Sphinx work is geared toward a universal SRS solution, or if it is focused on better individually "trained" programs.
I would be happy to share my idea with someone who has expressed some interest in a business partnership.
No pre-commitment necessary. ;-)
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I am interested in knowing if the Sphinx work is geared toward a universal SRS solution, or if it is focused on better individually "trained" programs.
I would be happy to share my idea with someone who has expressed some interest in a business partnership.
No pre-commitment necessary. ;-)
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Well as I mentioned before, Sphinx is a developers tool. You tailor it to address a problem. It does a good job in that it can do almost anything related to converting speech to text whether it be acting behind a asterisk server, voice control system to control your computer, dictation to convert whatever you say text, voicexml implementation, voice identification, etc...
Chances are, what you want to do, can be done. Its just a question of how well of a job it can do accuracy wise. Sphinx works out of the box, just not great for larger stuff for commercial projects. That takes time and effort. Lots. There are actually a number of companies that use sphinx. At least thats my impression of using sphinx so far.
I kind of ignored the part about universal SRS solution and individually trained programs since I am not sure what you mean by that.Go more into detail what you mean by universal SRS solution and individually trained programs. Almost looks like you mean speaker independant or dependant acoustic models?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
My "big" thought was to get a large number of folks to "train" the SR software for their individual use (this would be accomplished by making the software available to them at no charge). In exchange, they commit to forwarding their (vocabulary(?)) to us.
The program would be designed so that these individuals could easily record their own voice; perhaps from their cell phone calls? And the software on the cell phone would be designed to make it very easy for the person to take a few minutes when they have the time, to go back over what the program thought they said, and confirm and/or correct it.
I'm presuming that many thousands of these "vocabularies" gathered together would be the linchpin in creating a truly universal speech recognition software.
This is simply a thought which occured to me. as I said, I'm a marketer/entrepreneur, not a scientist. I would not be a bit surprised if the notion were complete hogwash. But it seemed worth asking someone who knew. What's the verdict, Elias?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
It is actually a good idea that is already being worked on by the open source project called voxforge (www.voxforge.org).
Their goal is to make an free open source acoustic model. They have a audio collector that works online and through telephone to collect speech. What happens is it gets you to read stuff online into the java applet or over the phone.
It just has a few hundred people with a combined ~50 hours of speech data. Though getting people to donate their voice/time seems to be difficult. Maybe it would be helpful to advertise that in donating your voice, speech apps you use will be customized to work for you and be more accurate overeall.
So, you and everyone reading this forum , go to voxforge.org and spend 10-20 mins donating your voice :>
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Greetings all,
I am a marketer, not a programmer/engineer, so please excuse in advance...
where can I find information about Sphinx which is written in layman's language?
where can I discover a listing of existing projects so that I can determine if anyone is already doing what I am interested in?
does anyone know if there are forums which help people like me (marketer/entrepreneur) to find engineer partners?
Thanks Elias.
Interesting.
It sounds as if there is not much energy (money) pushing the effort forward.
I wonder if a profit-incentivized model might be able to move this notion along much faster.
Also, it sounds as if this program does not allow for the individual to "train" the program... I will go and look beore I do any more guessing.
Thanks for the heads up.
Thank you, Elias.
I am interested in knowing if the Sphinx work is geared toward a universal SRS solution, or if it is focused on better individually "trained" programs.
I would be happy to share my idea with someone who has expressed some interest in a business partnership.
No pre-commitment necessary. ;-)
Thank you, Elias.
I am interested in knowing if the Sphinx work is geared toward a universal SRS solution, or if it is focused on better individually "trained" programs.
I would be happy to share my idea with someone who has expressed some interest in a business partnership.
No pre-commitment necessary. ;-)
By SRS you must mean speech recognition system.
Well as I mentioned before, Sphinx is a developers tool. You tailor it to address a problem. It does a good job in that it can do almost anything related to converting speech to text whether it be acting behind a asterisk server, voice control system to control your computer, dictation to convert whatever you say text, voicexml implementation, voice identification, etc...
Chances are, what you want to do, can be done. Its just a question of how well of a job it can do accuracy wise. Sphinx works out of the box, just not great for larger stuff for commercial projects. That takes time and effort. Lots. There are actually a number of companies that use sphinx. At least thats my impression of using sphinx so far.
I kind of ignored the part about universal SRS solution and individually trained programs since I am not sure what you mean by that.Go more into detail what you mean by universal SRS solution and individually trained programs. Almost looks like you mean speaker independant or dependant acoustic models?
I'll just come right out with it Elias.
My "big" thought was to get a large number of folks to "train" the SR software for their individual use (this would be accomplished by making the software available to them at no charge). In exchange, they commit to forwarding their (vocabulary(?)) to us.
The program would be designed so that these individuals could easily record their own voice; perhaps from their cell phone calls? And the software on the cell phone would be designed to make it very easy for the person to take a few minutes when they have the time, to go back over what the program thought they said, and confirm and/or correct it.
I'm presuming that many thousands of these "vocabularies" gathered together would be the linchpin in creating a truly universal speech recognition software.
This is simply a thought which occured to me. as I said, I'm a marketer/entrepreneur, not a scientist. I would not be a bit surprised if the notion were complete hogwash. But it seemed worth asking someone who knew. What's the verdict, Elias?
Hi Kurt
It is actually a good idea that is already being worked on by the open source project called voxforge (www.voxforge.org).
Their goal is to make an free open source acoustic model. They have a audio collector that works online and through telephone to collect speech. What happens is it gets you to read stuff online into the java applet or over the phone.
It just has a few hundred people with a combined ~50 hours of speech data. Though getting people to donate their voice/time seems to be difficult. Maybe it would be helpful to advertise that in donating your voice, speech apps you use will be customized to work for you and be more accurate overeall.
So, you and everyone reading this forum , go to voxforge.org and spend 10-20 mins donating your voice :>