Random Thoughts from coffee

Tom Anderson tanderso at oac-design.com
Fri May 21 20:13:52 CEST 2004


From: "Tom Allison" <tallison at tacocat.net>
> 10 might be too much, I was thinking of 5 eventually...  But it is my
> time at stake!  :)

definitely ham, probably ham, no clue, probably spam, definitely spam

I don't see too much value.  It was a long time before I saw any value in
"unsure" though.  In fact, I don't really use "ham, unsure, spam", I
basically use "ham, probably spam, spam".  The middle one is where I check
it harder just in case, but it's still almost always spam.  I tend to only
glance through the spam box and largely rely on other people telling me I
haven't responded to determine if I really got a false positive.  Hasn't
happened yet.  And I haven't missed any events or anything, so I'm pretty
sure there haven't been any.  When I get a ham, bogofilter has high
confidence that it is a ham.  You don't really need a "probably ham" box,
since you shouldn't be getting too many false negatives to simply drag them
to a folder for correcting later.  Basically, adding the third "unsure"
level has simply permitted me to lower my ham cutoff to where it should
really be, but still give myself a safety buffer to ease my worrying about
any false positives.  Adding another one or more levels just makes more work
and doesn't really solve anything.

For better false positive scanning (which isn't really needed as far as I
can tell), I would use Apple's vector method that we were discussing to
seperate all spams into categories.  This way, you could reject categories
like "viagra", "mortgage", etc., without even thinking about it, but maybe
look through "cooking" or "gardening" a little closer if maybe those are
subjects you sometimes discuss.

Tom




More information about the Bogofilter mailing list