Bayesian Teaching

[Jason Frisvold <xe...@gmail.com>]

Hi all,

I have a question regarding the bayesian filter, specifically the
learning function.  I use the sasql plugin for Squirrelmail which
creates "Learn Spam" and "Learn FP" folders.  I also have a procmail
script that moves spam to a Spam folder.

I was thinking about combining the Spam and Learn Spam folder. 
Probably create "Spam" and "Not Spam" instead, making it easier for
the users.

So, now on to my question.  Is it bad to continue feeding the same
spam to sa-learn?  The script that runs sa-learn also deletes the
spam/ham afterwards, but the chances of the same spam arriving again
are high, so I figure sa-learn will continue getting copies of the
same spam.  Will this cause any problems with the filter?

Thanks!

-- 
Jason 'XenoPhage' Frisvold
XenoPhage0@gmail.com

Re: Bayesian Teaching

[Jason Frisvold <xe...@gmail.com>]

On Wed, 27 Oct 2004 22:51:39 -0700, Robert Menschel <ro...@menschel.net> wrote:
> Hello Jason,

Hello :)
 
> No problem at all. I feed all spam from three domains into sa-learn
> for all three domains. Depending upon timing and other considerations,
> it's possible for a specific spam to reach domain A on Monday, domain
> B on Tuesday, and domain C the week after. All three emails are fed
> into sa-learn. The duplicates are identified by sa-learn and then
> ignored.

Awesome, that's EXACTLY what I was hoping...  Gotta make this spam
stuff as easy as possible for the users...  *grin*
 
> Bob Menschel

Thanks fot the info!

-- 
Jason 'XenoPhage' Frisvold
XenoPhage0@gmail.com

Re: Bayesian Teaching

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

Hello Jason,

Wednesday, October 27, 2004, 7:12:30 PM, you wrote:

JF> So, now on to my question.  Is it bad to continue feeding the same
JF> spam to sa-learn?  The script that runs sa-learn also deletes the
JF> spam/ham afterwards, but the chances of the same spam arriving again
JF> are high, so I figure sa-learn will continue getting copies of the
JF> same spam.  Will this cause any problems with the filter?

No problem at all. I feed all spam from three domains into sa-learn
for all three domains. Depending upon timing and other considerations,
it's possible for a specific spam to reach domain A on Monday, domain
B on Tuesday, and domain C the week after. All three emails are fed
into sa-learn. The duplicates are identified by sa-learn and then
ignored.

Bob Menschel