Program that uses Google's voice APIs now called Palaver. G+, videos, Github inside

[bboyjkang]

The open source GNU/Linux speech recognition program that uses Google's voice APIs on the back-end is now called Palaver. Google+ Community, installation video, and Github links are here:

https://github.com/JamezQ/Palaver

Palaver Speech Recognition Community on Google +:
https://plus.google.com/communities/...20902112738135

Information about Palaver
http://www.youtube.com/watch?v=a5-aolmt0OE

How to install Ubuntu Voice Recognition (Palaver by James McClain)
http://www.youtube.com/watch?v=pxom292XW_g

[bboyjkang]

Here are some responses by James McClain, the developer, to concerns from comments on a Reddit post:

"I also want to take some time responding to comments about this being a useless project.

I think most of the complaints are summed up in this comment by yepguy.

> The interesting part of this project is closed source and poses privacy issues, so it's of no use to the free software community.

There are valid points that bring up this comment, so I would like talk about some of them a bit.

> It needs Google, needs to be always online, privacy, etc.

Truth! This is a major flaw in the program currently, but the program has more to offer than just what is provided by google. It can amass a vast collection of dictionaries, able to recognize many many commands using fuzzy recognition. This program should be able to hook up to pocketsphinx just as well as google, gaining privacy and offline functionality for a possible cost of accuracy. This feature is something that has been planned for the start. Eventually, I would like people to hook up to any speech to text engine at all.

> The only thing interesting about this project is closed source anyway.

It's true, I have never programed a speech to text engine, but I have tried to make it clear from the start that I have not done that, that it uses google for that. I am sorry if that has been unclear to some of you. As for it being the *only* interesting thing, I disagree, I think providing fuzzy recognition dictionaries, and allowing people to install packages to get more functionality and voice commands is interesting.

> Why do you call it Palaver Speech recognition if it is doing natural language processing?

Because that is the functionality it provides to users. It may not actually be doing speech recognition, but to a user, it provides that functionality for him on his linux pc. I try to make it clear that I have not created a program to translate speech to text, but instead a program to provide the functionality of a speech recognition program, like siri, on linux."

http://www.reddit.com/r/linux/commen...ition_program/

[bboyjkang]

"Maybe, but that doesn't make this less interesting.

Google ran GOOG-411 for some *three years* to hone their voice recognition software on all sorts of phrases and accents. *Maybe* they have a version that would work offline, maybe there's source code that'd be interesting. But I'm guessing a large part of this involves a large database and a cluster.

Kind of like how there are probably interesting things in the Google Search source code, but a lot of it is just searching and updating a truly massive index. Having the source code wouldn't really give me "Freedom" in any sense that Stallman was after, unless I *also* have a multi-billion-dollar cluster at my disposal.

It means I can build an open source app that, given an Internet connection with which to talk to Google, can instantly have more reliable voice recognition than anything else out there -- which is going to continuously improve with time.

Actual open source voice recognition would also be interesting, but that doesn't make this concept *boring.*"

-SanityInAnarchy