[EMAIL PROTECTED] wrote: > You've lost me there. This is mail being analysed and adding to the picture > of your particular > blend of mail. Where is the 'incorrect bolstering' ? If it's mail, it gets > scored. You can't > dictate what ends up in spamdb. >
I don't consider the "pollution" I am referring to as my particular blend of mail, because it is from a select few people sending a *large* amount of repetitive personal messages. I am not trying to dictate what ends up in the corpus. These addresses are still whitelisted, and their "normal conversations" (that I don't redRe-list against) get into the corpus on a regular basis. I am addressing a specific situation that caused an imbalance of my corpus. An imbalance that caused false-negatives to get through the filter. I did *not* intentionally or pro-actively set out to do what I have been describing. I am not trying to control or micromanage the corpus. It was purely reactionary to a problem that presented itself after 3-4 months of use at this *specific* location. Consider this: I read the forums of other spam filters for ideas and known issues that admins have. Some admins hate Bayesian filters, because they claim they don't work. Why do you think that is? Isn't it just as simple as feeding it? Oh, its not? For instance: Why do we choose to not allow auto-reply messages add to the whitelist? Oh right, because we don't want to get spam in the corpus. But everyone that our users know must be OK, right? I mean they would never forward crap to our users, right? I mean, no one forwards junk mail (and misc spammy mail) around intentionally, right? To me, the problem is ultimately the same - it just might take longer for it to materialize into false-negatives. > Is this view too simplistic? Perhaps, but I try to keep things in simplest terms. The point I am trying to convey is that the problem I encountered is on a large scale and yet is of a very specific nature. As I've indicated, I've already been down the road of analyzing this problem and finding a solution that works. Take it for what it is, believe it if you want. I closely monitor the problem at this particular installation location, and I can attest to having eliminated the false-negatives as I described them. ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys -- and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ Assp-user mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/assp-user
