Greetings from the suddenly sprung northern US….and yes the fish are biting. First off, thanks for the opportunity to post here, and to tell you a bit about my project. I humbly ask your assistance with locating a semi-packaged? product(s) to begin experimenting with.

My goal? to provide a front end for a very complex CUDA (NVIDIA GPU) based piece of software designed for diagnosing catastrophic failures and performance degrades in enterprise software systems.

My requirements for this front-end would be to:
1) Be able to originate a telephone call based on a phone number input.
2) Monitor that telephone connection as a voice input stream for a known and limited set of commands.
3) Watch over and read in english “responses” from a text stream/file output from the CUDA app.
4) Send these “responses” not to the telephone, but to the web browser that originated the call request.
4) Ideally, the package would have built into it the sort of hooks that would allow one to change the “assistants” bots image.
- I’ve seen 3-D models rendered out and strung together that mimic “syllable facial expressions”. 
5) I would prefer to work in the open source world if at all possible, unless there is an existing product that you believe I could simply configure to achieve all of the above.

I’m professed to be highly ignorant of the current state of the technology you guys are into.
I even realize this question is pushing the limits of what one would probably consider a “chat bot”, but am hoping you’ll overlook that and point me to a starting place to work out this interface.

My thanks!
Greg

> superscriptjs.com

This sounds to me like a job for Rob Ellis’ (@rob_ellis) new “SuperScript”; since, one of his claims to fame is co-creating the phonegap.com mobile development framework.

Thanks Marcus.

Yes, I think superscript can do most of that.

I’m not really sure what is meant in #2. “Monitor Telephone conversations” Superscript (and CS, Rive, AIML) all deal with text streams and don’t have a native way to convert speech to text.

This could be accomplished by using another API. I would take a peek at nuance, google or maybe wit.ai. (now at facebook).

If you have any questions, ask away. chatbots.org is my favourite daily digest!

Thank you very much Rob and Marcus,

SuperScript installed and running on my development mac. It does indeed to seem to be a great running-start on what I have in mind!

I’m playing with some native Mac text to speech which will take a bit of XCode to get functioning (the native Samantha voice (which I believe to actually be Siri) is 400mb and pretty realistic. I also found DAZ Studio for pulling together a face and iClone to manage the animations.

The voice recognition piece seems to be the most troublesome. http://kaldi.sourceforge.net is probably close to what I’m after. It has several advantages, most notably the heavy lifting can be run on a GPU and is NVIDIA ready. Throwing a teraflop at the problem should overcome some inconveniences I’ve run into. The whole “phone” idea stemmed from my original experiments with a Siri Proxy server, which while it worked, the communication was unnatural, both because of the lag and having to “press to talk” and hear the dings as she listens and stops listening.

Accuracy of the voice recognition software I’ve toyed with is pretty dependent on the quality of the microphone, and sadly, the bluetooth mics I’ve tried are pretty lacking here. I’m hoping that throwing some computing power at it using Kaldi, the lower quality audio stream will function.         

Greg

at least you got a welcome i got alot of views but no replies on my   introducion post

welcome im new aswell