Spam in images
Tony L. Svanstrom
tony at moon.pp.se
Mon Aug 7 22:46:46 CEST 2006
On Mon, 7 Aug 2006 the voices made John G Walker write:
JGW> Can't comment on bogofilter, since I started with a couple of thousand
JGW> presorted spam and ham messages to train on, but my first Bayesian
JGW> filter - SpamBayes - only needed about thirty or forty messages before
JGW> it began to get the hang of my mail.
JGW>
JGW> I suspect any Bayesian filter will learn equally fast. I suspect the
JGW> speed is in the mathematics, not the software,
Well, it's not always that simple; if there's not enough information for the
filter to do its magic then it won't work.
Sure, if these e-mails are looking "too random", then the filters will be
updated so that they can find the patterns that they can't find today; but
there's a limit to how far you can take that without risking that unexpected
e-mails from new senders are classified as spam.
/Tony
--
/\___/\ /\___/\
\_@ @_/ \_@ @_/
.--oOO-(_)-OOo--------------------------------------oOO-(_)-OOo--.
| perl -e'print$_{$_} for sort%_=`lynx -dump svanstrom.com/t`' |
`---ôôô---ôôô----------------------------------------ôôô---ôôô---´
\O/ \O/ ©1998-2005 svanstrom.com \O/ \O/
More information about the Bogofilter
mailing list