Short answer: you train it. You give it some examples of spam and ham (that is, not spam), and it analyzes these messages. When a new message comes in, its contents are compared to the spammy and hammy tokens, and a spam rating between 0 and 1 is assigned (0 = "definitely not spam", 1 = "definitely spam"). You assign cutoffs for ham and spam ratings; anything in between is treated as unsure. When SpamBayes miscategorizes something, you tell it so, which adds new information to the training database. (This assumes you train on errors, which is only one possible strategy.)
Long answer: see http://spambayes.sourceforge.net/background.html and http://entrian.com/sbwiki/TrainingIdeas. > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of sachin chakote > Sent: Tuesday, February 08, 2005 3:40 AM > To: [email protected] > Subject: [Spambayes] How Spam Bayes determines no Of > Significant Words Inperticular Mail > > How Spam Bayes determines no Of Significant Words In perticular Mail? > > Sachin Chakote > Aftek Infosys Ltd. Pune _______________________________________________ [email protected] http://mail.python.org/mailman/listinfo/spambayes Check the FAQ before asking: http://spambayes.sf.net/faq.html
