SA Scoring

[Steve Campbell]

----- Original Message ----- 
From: "Dhawal Doshy" <dhawal at netmagicsolutions.com>
To: "MailScanner discussion" <mailscanner at lists.mailscanner.info>
Sent: Wednesday, May 17, 2006 11:07 AM
Subject: Re: SA Scoring


> Drew Marshall wrote:
>> On Tue, May 16, 2006 23:08, Scott Silva wrote:
>>> --[UxBoD]-- spake the following on 5/16/2006 3:09 PM:
>>
>>> As for a starter database, you can go to 
>>> http://www.fsl.com/support.html
>>> It should get your bayes working right away. It will still take time to
>>> build
>>> accuracy, but should get you started.
>
> I would recommend that you manage to create your own database from scratch 
> in a few weeks, rather than using the starter database forever.
>
> A starter is meant for just starting out, nothing more. The accuracy will 
> always be higher in a manually picked (your own) spam/ham learning as 
> compared to a started database.
>

I've always wondered about this point. As the spam we receive seems to run 
in particular spurts from particular spammers with specific content, the 
saved emails that I might keep to start up a new db file would appear to be 
outdated whenever I needed them again for priming the new db.

Granted, it would be more pertinent to use spam to my mailservers than to 
use a generic starter DB, but would I gain anything other than having the 
required 200 mails, no matter which set of emails I use? The 'seen' stuff 
may be unseen forever again.

Steve Campbell
campbell at cnpapers.com
Charleston Newspapers


> I could be wrong though, Matt can correct me here.
>
> - dhawal
> -- 
> MailScanner mailing list
> mailscanner at lists.mailscanner.info
> http://lists.mailscanner.info/mailman/listinfo/mailscanner
>
> Before posting, read http://wiki.mailscanner.info/posting
>
> Support MailScanner development - buy the book off the website!