Steffen Nurpmeso wrote in <20230221173236.ulzv1%stef...@sdaoden.eu>: |Χάρης Καραχριστιανίδης wrote in | <20230221033050.uf_w5%haris...@otenet.gr>: | ... ||Thank you, my problem is that even emails that I define as spam with \ ||spamspam have score 0.00 (as all other mail).. I trained bogofilyrt \ ||with about 450 emails till now which I marked as spamham (most of then \ ||and a few as spamspam). Does it need more training to show ratings \ ||other than 0.00? | |Then yes. :-)
The thing is Bayes counts words. It is pretty primitive. And usually nothing is wrong with words like mastu***tion or such, in fact it is pretty natural for at least the human species, but in some context it may be (super annoying or disgusting) spam. This is in fact why i removed the Heirloom spam code, even though it has its merits (word anonymization for example). Because now you can use whatever you want, and exchange the DBs that were trained over a long time in between whoever needs a trained Bayes spam checker. In real life i expect noone except myself to use this MUA alone. (Some old hands keep using Heirloom though; one aspect could surely be that they loose a trained Bayes.) Ciao, Hariskar! --steffen | |Der Kragenbaer, The moon bear, |der holt sich munter he cheerfully and one by one |einen nach dem anderen runter wa.ks himself off |(By Robert Gernhardt)
