Hello BJH & everyone else, on 04-Feb-2005 at 14:56 you (BJH) wrote:
> It seems to me that in simple terms the plugin assigns a 'score' to each > email, and dependent on the score allocated TB determines what ultimately > happens to that email. To be precise, by the classification of each mail the words in the message text get a score. BayesIt collects these words with an individual ranking in a database (something like a statement: "the word 'telephone' appeared in ham emails 15 times"). Over time it becomes clearer and clearer which words appear in spam email and which in ham email, and the bayes filter "weighs" each word and classifies the messages accordingly. > I've been 'training' or at least telling TB what is junk over the last few > days. Currently 107 spam mails against 170 genuine. As Marck told you, the filter needs input for training, and so far you have only very few genuine mails. You could mark your entire "sent mail" folder and classify the messages as ham to increase the dictionary for ham mails. -- Best regards, Alexander (http://www.neurowerx.de - ICQ 238153981) Stupidity got us into this mess - why can't it get us out? ________________________________________________ Current version is 3.0.1.33 | 'Using TBUDL' information: http://www.silverstones.com/thebat/TBUDLInfo.html
