Bayesian Filter for Slashdot Moderation?

Spam filters using bayesian methods [www.paulgraham.com] are becoming very popular (and accurate!). After learning a bit, they are capable of picking out spam pretty well. How about applying a bayesian filters to Slashdot posts and seeing what happens? The "teaching material" (old posts with scores) is readily available. Will a trained filter recognize a "5, Funny"? or a "3, Interesting"? Who is willing to build such a post judge?