The whole point of these are to beat Bayesian filters which use word
occurrence as a spam filtering policy. Basically (and I'm probably
getting this at least 50% wrong) is that a Bayesian filter would do
something like note: "All emails which have the word Viagra are
ALWAYS spam" so it would always filter out some of those spam.
I think the Bayesian filter doesn't have any info on words like
clarify or directorate (or at least they're rare in spam mails) so
the filter doesn't tag those as spam.
Adam
I believe the filter will actually lower the estimated probability of
a message being spam if it contains certain words.
Now they need to do some sort of 'word distribution' scanning to
catch the spammers. And the etymologists :)
Have fun,
Shay
--
=========================== Shay Telfer ================================
Perth, Western Australia Technomancer Chronopolis Quiz Day
Opinions for hire [POQ] Sun. March 14, Hyde Park Hotel
[EMAIL PROTECTED] fnord <http://chronopolis.sf.org.au/>
