Scanning Wikipedia to build a statistical model was an idea that one of our contributors came up with. We haven’t explored it or fleshed it out in more detail yet because that’s working towards the long-term goal of building our own termbase. For the immediate future, our priority is getting a working tool in an upcoming release of Firefox as fast as possible. Google spent 10 years building their termbase. We’re up for it, and we’d welcome your efforts along that line, but it’s unlikely to yield quick results.
That being said, feel free to take on that part of the project! You can start an Etherpad and begin specifying what is required to do the scanning and statistical modeling. I’m sure other interested developers will join in! Please let us know in #intellego if we can help you along.