On Sun, Jan 02, 2011 at 08:30:43AM -0800, Marc Perkel wrote: > Here's a wild idea that might prove a point. Create a set of meta > rules which is a combination of every set of two rules. > > meta COMBO_RULE1_RULE2 (RULE1 && RULE2) > describe COMBO_RULE1_RULE2 RULE1 and RULE2 > score COMBO_RULE1_RULE2 0.1 > > Then run stats to see if any of the combos produce interesting and > useful results. Then do 3 rule combos. > > I'm betting that new useful rule combos will be discovered > > Someone could write a perl script hat would generate the rules.
Um helloo. http://www.mail-archive.com/users@spamassassin.apache.org/msg61057.html
