RE: What don't we just use a spam filter for comments?

Sounds great! Are you going to use Naive Bayes or some kind of neural network? Or maybe different approach?