It's been well over two years since I last maintained my blog, and spam had gotten the better of it. I rigorously reviewed my Naive Bayes implementation, and sure enough, spotted some bugs.
Now that the positive spam classifier is reset, all comments should trigger a Turing code challenge, which with some additional complexity now hopefully further discriminates humans from malicious web-crawling bots.
While back in 2005 it was fun and challenging to write my own spam filter, the open source community has since truly opened up, with magnitudes of great implementations. Now I just have my fingers crossed that my simple classifier dismays most of those spammers.