Vanner Vasconcellos wrote: > A little bit off-topic, but.... > > Does someone knows if there is any academic research on fighting SPAM? > > Maybe in areas like "Artificial Inteligence" or "Knowledge Discovering"?
There's quite a few, the best I found was a bibliographic reference to all the research out there. Find it here: http://www.esi.uem.es/~jmgomez/spam/MLSpamBibliography.bib Note however that most of the research isn't very good - it focuses on training and testing based on PU1 corpus, which is old and pretty weak by today's standards, and they also have very poor corpuses for training non-spam results on (usually mailing list archives, which aren't representative of most people's everyday email). Having said all that, there's value in the research. The best technique seems to be Naive Bayes, though I'm investigating combining that with Boosting Trees. (This stuff is all part of the value-add on to SpamAssassin here at MessageLabs). Matt. ---------------------------------------------------------------------------- Bringing you mounds of caffeinated joy >>> http://thinkgeek.com/sf <<< _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
