> -----Original Message----- > From: Dallas L. Engelken > Sent: Monday, December 01, 2003 8:49 AM > To: [EMAIL PROTECTED] > Subject: Re: [SAtalk] Evil rules, popcorn, etc too much? > > > >> The FP rate for all of these is just about zero. > > > >Has anyone run these on a large corpus to see how it goes? > > > >(I've seen no mass-check-style output for these rules.) > > > > here are the rules on my corpus... about 5k spam and ham.. 10k total. >
i updated my popcorn, blackhair and weeds / weeds2 rules from the 11/13 ruleset, and re-ran the corpus tests (after removing 30 emails, see below)... here they are.. http://engelken.net/masses/testrule.NEW_BLACKHAIR.txt.out http://engelken.net/masses/testrule.NEW_POPCORN.txt.out http://engelken.net/masses/testrule.NEW_WEEDS_TEST_2.txt.out http://engelken.net/masses/testrule.NEW_WEEDS_2_TEST_2.txt.out - 0 total hits on the ham corpus for the popcorn tests. - 7 total hits* on the ham corpus for the Blackhair tests. * see hard_ham/00189.c69e4af5bfa5a1bcb403eefe112e5e45 - it takes 4 of the hits - 0 total hits** on the ham corpus for the weeds and weeds 2 tests. ** the previous weeds and weeds 2 tests had alot of hits on the ham corpus, but *ALL* of the hits were all related to the 30 emails on the <sprocket.lockergnome.com> mailling list that you can find in the public hard_ham corpus. [EMAIL PROTECTED] masses]# grep J_WEEDS_C ham.log | wc -l 30 [EMAIL PROTECTED] masses]# grep J_WEEDS_C ham.log | grep lockergnome | wc -l 30 I removed the following messages from the ham corpus and ran the weeds tests again.. http://engelken.net/masses/offending_weeds_email.txt The filenames you see there can be matched against what you find in hard ham.. http://spamassassin.org/publiccorpus/20030228_hard_ham.tar.bz2 Dallas ------------------------------------------------------- This SF.net email is sponsored by: SF.net Giveback Program. Does SourceForge.net help you be more productive? Does it help you create better code? SHARE THE LOVE, and help us help YOU! Click Here: http://sourceforge.net/donate/ _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
