[SAtalk] Spam confuses bayes auto_learn

2004-01-15 Thread Gunther Heintzen
Hello In the last time we often got spam emails containing random word lists. example: nagasaki gazette rood cod dual gradate clifford babysitting newel latin adenoma francisco aspheric express atavism guardia photo section copperas ... Some of them are recognized as spam some as

Re: [SAtalk] Spam confuses bayes auto_learn

2004-01-15 Thread Matt Kettler
At 11:14 AM 1/15/04 +0100, Gunther Heintzen wrote: X-Spam-Status: No, hits=2.6 required=3.9 tests=FORGED_HOTMAIL_RCVD2, HTML_MESSAGE autolearn=ham version=2.61 It should be autolearn=no because hits=2.6 ist beetween 0.1 and 12.0 Autolearning is not based on the normal message score, it's based