>>pg+(OP)
3 words: Bayesian Comment Filter. Just does the opposite of what the spam filter does. Use the corpus of great comments from the past to find great comments of the present.
I'm only half joking. Fundamentally, the thread is about a filtering system.
>>b_emer+A4
Perhaps you could train a Bayes classifier on users' voting patterns (beginning with a corpus of good and bad comments) and use that to a) decide how to weight users' votes and b) classify further comments as good or bad. (This makes no attempt to classify comments by their content--that'd be done in the construction of the original corpus.)