Martijn Grooten schreef: > Dear Amedee, others, > > Thank you for your email. > > To answer your question: if I understand it correctly, Spambayes > takes an email message (header + body) as input and outputs either > 'spam' or 'ham'. (I know many products actually output a likeliness > percentage, but by setting a threshold this has the same effect.)
Spambayes outputs a number between 0.00 and 0.99. > This does not necessarily have to be a problem: we could possibly > install Spambayes on the server which redistributes the email and, > once an email is redistributed, make it check whether it thinks the > email is ham or spam and store that in the database. Note, however, > that for various reasons, we do not 'teach' the filters by providing > them with end user feedback. Teaching is essential for Spambayes. It simply cannot work if it doesn't get a statistically relevant corpus of ham and spam to train on. Without training, all emails will get a score of 0.50. I don't want to speak on behalf of the developers, but I'm afraid that Spambayes falls outside the rules of your test. Kind regards, Amedee Van Gasse _______________________________________________ [email protected] http://mail.python.org/mailman/listinfo/spambayes Info/Unsubscribe: http://mail.python.org/mailman/listinfo/spambayes Check the FAQ before asking: http://spambayes.sf.net/faq.html
