[bogofilter-announce] bogofilter-0.14.0 - New Current Release
Boris 'pi' Piwinger
3.14 at logic.univie.ac.at
Wed Jul 23 14:21:09 CEST 2003
David Relson wrote:
> * Default wordlist mode is single, combined wordlist.
> File wordlist.db contains all spam and ham tokens.
Before:
> [3.14 at pi ~/.bogofilter]$ bogoutil -w . .MSG_COUNT
> spam good
> .MSG_COUNT 838 495
> [3.14 at pi ~/.bogofilter]$ du -sk .
> 4340
After:
> [3.14 at pi ~/local/bogolists/.bogofilter]$ bogoutil -w . .MSG_COUNT;du -sk .
> spam good
> .MSG_COUNT 838 466
> 4132 .
Surprising to me is that there is no significant size effect.
Also:
spam good
.MSG_COUNT 604 410
Added 46 ham mails and 213 spam mails to the database.
spam good
.MSG_COUNT 815 456
604+213=817>815, right?
Added 9 ham mails and 24 spam mails to the database.
spam good
.MSG_COUNT 837 465
815+24=839>837, again two missing.
Added 1 ham mail and 3 spam mails to the database.
spam good
.MSG_COUNT 838 466
837+3=840>838, again two spam, suspicious.
Let me try to train with the saved messages:
> [3.14 at pi ~/local/bogolists]$ bogofilter -d .bogofilter -sM<bogominitrain.spam
> Segmentation fault
> [3.14 at pi ~/local/bogolists]$ bogofilter -d .bogofilter -nM<bogominitrain.ham
I found two messages which cause a segfault. The rest still
does that too:-(
http://www.logic.univie.ac.at/~3.14/bogosegfault.zip
pi
More information about the Bogofilter
mailing list