Lectures are over for uni at the moment, so it’s that time of year again when I start doing work. Inspired from watching Iron Man the day it came out in NZ I decided to build my own voice powered bot.
I’m too lazy to write my own speech recognition engine, so I started using SAPI 5.3. But I’ve switched over to using the .NET 3.5 Speech library at the moment.
Image may be NSFW.
Clik here to view.
The hard part now is defining the formal grammar, as all it does now is check for “computer email” and opens up webmail for me. But given that dynamic grammars can hook directly into databases the possibilities are incredible.
Unfortunately I need to grab another 900MB SDK to quickly build these up. Otherwise I’d have to work with the w3c spec by hand. So I’ll leave it on hold for a bit until my XPS 1530 arrives (I don’t have 900MB of room left on my hdd…)
Also I need to think about how to map the resulting accept state into the list of actions to perform. Since currently I just know that I have an accept state, and the string, but what I really want to have is the list of states that I had to pass through to get there.
i.e it’s no good knowing that “computer fetch images from my documents”, what I want is something like:
computer -> fetch -> images -> from -> my documents
Not computer -> fetch -> images -> from -> my -> documents
as I’m too lazy to do a double parse.
That way I quickly build up some sort of structure for scripts, so that it can look for “computer fetch $1 from $2″ and just pass along the required variables.
Though since a week is a long time to wait I’ll write a speech to action program for unique speech strings on wednesday.
So to keep Mathew happy, since it can run an arbitary url you can in fact have “computer order ice cream” and it’ll work.
