Bugs item #1199978, was opened at 2005-05-12 04:33
Message generated for change (Comment added) made by anadelonbrin
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=498103&aid=1199978&group_id=61702

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: Outlook
Group: 1.0.1
>Status: Closed
>Resolution: Duplicate
Priority: 5
Submitted By: Karl Vollmer (vollmerk)
Assigned to: Tony Meyer (anadelonbrin)
Summary: Requires Owner on Public Folder to Score Spam

Initial Comment:
In an attempt to make it easier to train Spam Bayes for
all of our customers we created a Public Folder in
exchange that contained a large number of known Spam.
What we would like to be able to do is when installing
Spam Bayes for a new user simply point it at this
Public Folder to train for "Bad Messages" unfortuantly
we ran into a problem.

If the user has read only rights to the Public Folder
when you point the Spam Bayes configuration agent at
it, it appears to process the messages but then comes
back with a "0 Spam Messages Found" If you set the user
at a owner on the folder and try again it works
perfectly. Unfortuantly setting every single one of our
users as a "Owner" really isn't an option (we don't
want them deleting messagesf from it etc). 

It would be nice if Spam Bayes didn't require owner of
the folder in order to score spam messages. 

----------------------------------------------------------------------

>Comment By: Tony Meyer (anadelonbrin)
Date: 2005-12-05 21:54

Message:
Logged In: YES 
user_id=552329

Dupe of:

[ 903440 ] Fail to filter public folder
http://sourceforge.net/tracker/index.php?func=detail&aid=903440&group_id=61702&atid=498103

----------------------------------------------------------------------

Comment By: Tony Meyer (anadelonbrin)
Date: 2005-05-12 13:03

Message:
Logged In: YES 
user_id=552329

In general, I wouldn't recommend this approach for a number
of reasons.  The first is that it's really best to only
train on mail that you receive - there are clues (addresses,
received headers, etc) that will not be correct for each
user, and the idea is to train on spam unique to each user.
 Secondly, this is likely to lead to an imbalanced database
(more spam than ham) unless you can be sure that the users
will train on similar amounts of ham to start with (and
imbalanced databases are a bad idea).  Thirdly, generally a
small database is usually better than a large one, and this
may lead to a large database.  See
http://entrian.com/sbwiki/TrainingIdeas for more information.

If you do decide to do this anyway, it would probably be
easier to simply train a database on this mail and give the
database files to each user to start with rather than
getting them to do the training in Outlook.

As for the problem, the owner privileges are required to
score messages because the spam field is modified/created. 
If you turn off writing the spam field (not really a good
idea) as described in the configuration guide, this may work.

If you train without rescoring the messages, does that work?

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=498103&aid=1199978&group_id=61702
_______________________________________________
Spambayes-bugs mailing list
[email protected]
http://mail.python.org/mailman/listinfo/spambayes-bugs

Reply via email to