mailbox classification not working right
Barry Gould
BarryGould at PennySaverUSA.net
Tue Feb 4 19:52:01 CET 2003
Hi,
With Bogofilter 0.10.0, using Robinson-Fisher ternary:
I just fed a mailbox full of spam to bogofilter -s, and, although it did
register the mail, I do not believe it is doing it correctly.
I believe it is treating the entire mailbox as only a few messages. (more
evidence below)
The reason I am complaining about this is that it is not doing a good job
registering the messages as spam. If I take a single "Unknown" (spamicity
0.36 or so) message from the mailbox and register it, the Spamicity
immediately goes to 1.0000, but if I register the mailbox, it only changes
VERY slightly.
I read part of the recent "mailbox classification" discussion; I understand
Formail may be recommended in the future, but it seems like it needs to be
recommended for 0.10.0 as well.
BTW, I registered all my corpora several months ago this way; I wonder now
if it didn't work right then either.
Thanks,
Barry
more evidence:
"webmaster" is a linux mailbox which contains 26 spam messages, some of
which were unknowns or false negs.
# grep Subject\: webmaster|wc -l
26
# grep From\: webmaster|wc -l
26
# bogoutil -w .bogofilter/ .MSG_COUNT
spam good
.MSG_COUNT 10185 32689
# bogofilter -s < webmaster
# bogoutil -w .bogofilter/ .MSG_COUNT
spam good
.MSG_COUNT 10190 32689
More information about the Bogofilter
mailing list