Nick Simicich wrote:
One thing about Bayesian spam filters (I use and contribute to bogofilter, which has performance as a goal, see http://bogofilter.sourceforge.net) is that they are typically tunable.
So what? a) That takes work. I don't want to work at it.b) That occurs aftyer false positives have happened. It's unacceptable to me that they *ever* happen.