[denniskane]

I have a concern that is somewhat related to this issue. I've been wanting to play around with doing NLP in the client using the SpeechRecognition API available in Chrome. It typically gives pretty good results for simple recognition purposes, but there is not yet any way to specify an arbitrary grammar for the backend service to use. This is a big deal when someone wants to be able to recognize a name spelled like "Bobbie" rather than "Bobby". The W3C Web Speech spec currently allows for setting arbitrary grammars, but it isn't implemented. I'm just saying it would be cool if some of you AI geniuses at Google could help me out on that front.

You can play around with the dumb little AI app that I've been working on that does a little bit of NLP in the browser. Check out this site with Chrome: https://yc-prototype.appspot.com. His name is Bertie. You should see his face on the bottom of the page. The site works like an OS in your browser.