Gregory Mokhin a écrit :
Looks like same spammers that had used geocities before now send
messages with new urls (an excerpt):
****
Free check-up details review with our approved expert.
http://if.jlp.forwardthebest.com/n4j/
message to oz, saying if he lilyhanded did not let them in self-politician
ruby port to see him at once they
****
A question: is it actually useful to train SA on these messages?
Doesn't the garbage after the url just poison the bayes db?
You should train SA on all errors. if the words appear in lot of spam,
then you get the training. if it happens in a lot of ham, it won't
affect the results.
and after all, bayes is not the only test in SA.
-------------------------------------------------------
This SF.Net email is sponsored by:
Power Architecture Resource Center: Free content, downloads, discussions,
and more. http://solutions.newsforge.com/ibmarch.tmpl
_______________________________________________
AMaViS-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/amavis-user
AMaViS-FAQ:http://www.amavis.org/amavis-faq.php3
AMaViS-HowTos:http://www.amavis.org/howto/
