Learning from attachments

[Mike McMullen <ml...@loanprocessing.net>]

Hi all,

I currently have spam delivered to an IMAP folder as an attachment
via MailScanner. 

If I feed the IMAP folder to sa-learn using the --mbox flag, does it know
to look at the attachment as the item to learn? Or am I screwing up here?

Thanks,

Mike

Re: Learning from attachments

[Mike McMullen <ml...@loanprocessing.net>]

From: "Matt Kettler" <mk...@evi-inc.com>
Subject: Re: Learning from attachments


> At 02:36 PM 3/1/2004, Mike McMullen wrote:
> >If I feed the IMAP folder to sa-learn using the --mbox flag, does it know
> >to look at the attachment as the item to learn? Or am I screwing up here?
>
> Option B)
>
> SA-learn won't know to learn from attachments generated by MailScanner.
> Mailscanner generates it's own formats which are a bit different from the
> ones SA generates.
>
> You'll need to find some way of using an intermediate tool to extract the
> attachments and feed those to sa-learn. You might be able to abuse mutt
for
> this task, but I'm not really up-to-speed on command line tools for mime
> parsing.
>
>

Thanks, Matt. I was afraid of that. I've probably fed about 100 spams to
sa-learn before thinking to ask this question. Has this really screwed up
the
bayes database?

Also, if I change MailScanner to not use attachments but just deliver, would
that work with feeding the IMAP folder to sa-learn?

Thanks!

Mike

Re: Learning from attachments

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

At 02:36 PM 3/1/2004, Mike McMullen wrote:
>If I feed the IMAP folder to sa-learn using the --mbox flag, does it know
>to look at the attachment as the item to learn? Or am I screwing up here?

Option B)

SA-learn won't know to learn from attachments generated by MailScanner. 
Mailscanner generates it's own formats which are a bit different from the 
ones SA generates.

You'll need to find some way of using an intermediate tool to extract the 
attachments and feed those to sa-learn. You might be able to abuse mutt for 
this task, but I'm not really up-to-speed on command line tools for mime 
parsing.