Regex filters => Bayesian tokens?

[pna.lists]

Which is the easiest way to convert the results of checking the e-mail
using several regex filters into additional Bayesian tokens?

The SpamAssassin rules
(http://spamassassin.apache.org/tests_3_1_x.html) and the rulesets at
http://www.rulesemporium.com/rules.htm are useful, but the traditional
SpamAssassin fixed cumulative score approach leads to false positives.