On 31 Oct 2002, at 7:55, Sean Brunnock wrote:
> SpamAssassin has arbitrary rules for determining if an
> email message is spam. I think the fact that SpamAssassin
> tagged a french soccer newsletter as porn would indicate
> that.
Indeed it does have somewhat arbitrary rules. SpamAssassin is basically
a probabilistic spam catcher. In my experiece, running just out of the
box [that is, with my not 'helping' it], it is getting about 1.5% false
positives [for me, that's two or three or four messages a day that end up
erroneously in my spam folder] and about 1% false negatives [oddly, it is
about the same: two or three or four spams a day that make it into my
real mailbox.
When I say "out of the box", that means that my *FIRST* rule in my
incoming mail filter is "if SA doesn't like it, put it in the spam
folder". Some mailing lists I'm on *DO* look like spam [e.g., there are
some that are advertiser sponsored, and so it is not surprising that SA
sees the 'ads' in the list-messages]; some of the spam I get is legit
[that is, announcements from companies that I *asked* to send me info
about product updates].
If I cared enough [and at the current FP rate, I don't] I'd put my
mailing list filing and other such filters *ahead* of the SA filter, and
that'd take my FP rate down very low [maybe a FP a week, if that much].
One of the guys at work does this [has the SA filter as the very last one
in his filter set] and he says that he can't remember the last FP he got.
On your original comment, you're probably right: I'd guess that SA's very
cleverly tuned filters and weights are very highly biased toward spam-in-
english and might well not be *NEARLY* so good with traffic [spam or not]
in other languages. That's fixable, of course -- you can go in and tune
the filters and tweak it up to understand French or German or whatever.
It's just a big Perl program and you could join the project to help out
and get SA to work better in other languages [if, indeed, it really does
not].
But IMO the right way to be using SA is *NOT* the way I"m using it [and
the way you apparently are, also], as your first-cut preliminary incoming
email filter, but rather as the *LAST*. That is, set up your filters so
that they deal with your mailing lists [most of us file mailing-list mail
into appropriate folders]; set up some filters to accept "spam you REALLY
DID request"; and the only when you get to the very _bottom_ of your list
of filters, when you're about to dump the incoming message into your
inbox, THEN do the SA filter. I can't say about how it'll work for folks
in France, but I know that for email-in-English it is really very good.
/Bernie\
--
Bernie Cosell Fantasy Farm Fibers
mailto:bernie@fantasyfarm.com Pearisburg, VA
--> Too many people, too few sheep <--
Follow-Ups:
References:
|
|
