I worked on a similar project a decade ago written mostly in Perl with the frontend in PHP (hey, it was 2004, folks!). Just checked and I still even have the old courtbot.com domain I registered for the project.

I suspect you'll find pretty quickly that there's a limit to how far regular expressions or similar techniques can take you if you want to normalize and reference precedents and make sense of cases. That's why Lexis and Westlaw pay actual attorneys considerable sums to summarize cases, and why they can still command such princely subscription fees even in 2014. But analytics might be interesting. A family member is a judge, and her judicial office keeps track of how many cases she decides per month, how many reversals she receives, etc. I don't know if those are made public -- certainly I'm not aware of any project to do it across a large data set, and I wish you luck with it.

You're definitely right about case/precedent information, but what we've found is that there's a whole other world of info that can be neatly organized with a lot of crunching, and a small bit of manual manipulation.

The big guys chasing this are highly focused almost entirely on lawyers, in the context of providing them case analysis tools. We've found a bit of a different niche which doesn't need as much fidelity/granularity to the information, but needs it nonetheless.

In any case, I'd love to chat about your experience, even if a decade old. Can I PM you?

Sure, happy to chat! What you're doing seems interesting, especially if you're not targeting the lawyer/case research market. My email address is in my HN profile. Though I am working nonstop on http://recent.io/ right now. :)