VoxForge
VoxForge
Here is a proof of concept app (http://www.voxforge.org/testApp.html) using the WAMI toolkit to create a game to collect speech.
It is a crude attempt at 'speechifying' the Free Rice word definition game. There is only one definition and no scores are kept.
Do you think that something like this would encourage more people to donate speech? How would you improve it?
If the game is successful, we could modify it to use a VoxForge server for recognition so that we could collect speech in a more engaging way.
Do you have any other game ideas? If so, post the WAMI toolkit javascript code in this thread, and I can set it up on the VoxForge server for others to try (or post a link to your server) - the toolkit is very easy to use.
thanks,
Ken
P.S. The WAMI toolkit was originally discussed in this thread: Using Games as a Way to Get Users to Submit Speech.
--- (Edited on 9/16/2009 11:49 pm [GMT-0400] by kmaclean) ---
--- (Edited on 9/16/2009 11:50 pm [GMT-0400] by kmaclean) ---
Very interesting!
To improve it, I would do the following (bear in mind that I am not aware of any restrictions of the toolkit):
Make it possible for users to click a button to activate recording and to stop recording when the user turns silent. Because I feel that it is a bit awkward for a user to press and hold a button, especially since it would be possible to make a version where the user could simply click the proper answer.
It would be even more convenient to the user if he could strike a key on his keyboard to start recording (or if this would be an alternative option (e.g. the spacebar). That way, speaking would have a true advantage over using a mouse (not having to move the pointer). That's a small advantage, but I think any advantage is necessary, since for many people a simple click-variant is just as attractive.
--- (Edited on 9/19/2009 5:06 pm [GMT-0500] by Robin) ---
All of this is possible with the Toolkit, but has not been documented because of a Safari bug which for some reason will not allow Java and Javascript to communicate via LiveConnect when the Java is from a 3rd party site. The WAMI server is open-source (http://code.google.com/p/wami/), we'd love your help improving it!
--- (Edited on 1/5/2010 8:43 am [GMT-0600] by Visitor) ---
ecret has created Speechle an online quiz engine that uses speech recognition.
It uses javascript and flash on the front-end. Not sure what he is using on the back end (does not seem to be using the WAMI toolkit), but based on the questions he asked on the VoxForge forums, it is likely Sphinx-based.
He also includes a Javascript API to "enable your web apps with speech recognition and text to speech with just a few lines of code."
--- (Edited on 1/7/2010 1:44 pm [GMT-0500] by kmaclean) ---