Hello Watcher & everyone else, on 17-Dez-2004 at 18:03 you (Watcher) wrote:
[ BayesIt ] > It starts out with it's basic knowlege about spam I don't know of any "desktop" spam filter with Bayes analysis that comes with a pre-defined dictionary. You have to teach any Bayes filter (not only BayesIt) what is good (ham) and what is bad (spam) mail continously - its accuracy will increase over time and you have to do less training in the long run and only need to occasionally re-classify one or the other mail. The nice thing about BayesIt is (or theoretically, would be) that you can immediately train it from within TB - mark everything in your "Sent" folder and classify it as "not junk" (server-side commercial spam filters can often auto-learn from sent messages, too). And you could download "collections" of spam mails from the internet, import them to TB and train BayesIt with those (similar game, select them all and classify as "junk"). However, this won't fit your very own personal "spam profile" (one person may consider the weekly offers from his/her freemail provider valuable information, the other is annoyed by it and classifies it as spam), I believe importing such collections actually weakens the filter because it is not precise enough to catch the mails that you, and only you, consider spam. I had the same problems with BayesIt as Roelof had - I teached and teached it with the spam that went to my Inbox, but it never seemed to catch up. With different mail aliases (7) and accounts (2), I get the very same spam message repeatedly, yet BayesIt didn't recognize them, not after re-classifying the first, the second, the third... I don't know whats wrong with it, and why it works fine for others - for me, it didn't. Having used SpamPal + Bayesian plugin, PopFile & lately K9, they all performed a lot better even after the shortest training periods. -- Best regards, Alexander (http://www.neurowerx.de - ICQ 238153981) using TB! v3.0.2.10 Home on Windows XP Pro Service Pack 2 A learned blockhead is a greater blockhead than an ignorant one. -- Benjamin Franklin ________________________________________________ Current version is 3.0.1.33 | 'Using TBUDL' information: http://www.silverstones.com/thebat/TBUDLInfo.html
