Keeping old spam around

[Burton Windle <bw...@fint.org>]

Once a piece of spam has been shown to sa-learn as spam, is there any real 
use in the user keeping it around? I know that sa-learn will skip over it 
if shown it again, so I'm thinking "no" (unless you lose your bayes db and 
want to train it again quickly, but that's what backups are for).

-- 
Burton Windle                           bwindle@fint.org

Re: Keeping old spam around

[Robert Menschel <Ro...@Menschel.net>]

Hello Burton,

Monday, August 22, 2005, 1:29:48 PM, you wrote:

BW> Once a piece of spam has been shown to sa-learn as spam, is there any real
BW> use in the user keeping it around? I know that sa-learn will skip over it
BW> if shown it again, so I'm thinking "no" (unless you lose your bayes db and
BW> want to train it again quickly, but that's what backups are for).

In addition to having it available to rebuild your database, the next
best use for old emails is as a corpus, so you can run your own
mass-checks against them.

This comes in very handy for helping the SA development team identify
which prospective rules are most promising, and is used during the
release cycle to determine the scores for rules with each new major
release.

Bob Menschel

Re: Keeping old spam around

[Matt Kettler <mk...@evi-inc.com>]

Burton Windle wrote:
> Once a piece of spam has been shown to sa-learn as spam, is there any
> real use in the user keeping it around? I know that sa-learn will skip
> over it if shown it again, so I'm thinking "no" (unless you lose your
> bayes db and want to train it again quickly, but that's what backups are
> for).
> 

You would be correct. The only point in keeping it around would be to rebuild
your bayes DB, or if you wanted to do some kind of further analysis on the
message yourself.

Either way, SA has no direct need for you to keep your messages around to make
bayes work.