[Spambayes] Are they learning?

Neil Schemenauer nas at python.ca
Sun Feb 16 07:12:51 EST 2003


Bill Yerazunis wrote:
> Actually, I wouldn't worry about it.  That format still has 98% or so
> of the tokens intact, and the phrasing is still heavily spam.

Spambayes with my DB still finds lots of clues:

   '*H*' 0.100221799828
   '*S*' 0.954253123262
   'skip' 0.0256108900427
   'task' 0.0364772269246
   'header:Message-id:1' 0.0602125241541
   'content-type:text/plain' 0.254441488712
   'chance' 0.277902899943
   'want.' 0.309419424496
   'going' 0.371615649759
   'try' 0.384359911306
   'skip:e 10' 0.63826549096
   'link' 0.642099485564
   'proto:http' 0.662211682949
   'url:www' 0.68175741844
   'online' 0.725029235286
   'visit' 0.726863723599
   'ready' 0.738706256842
   'url:com' 0.752211285796
   'here' 0.771157119995
   'below' 0.819844718374
   'now!!!' 0.844827586207
   'url:tropicalpills' 0.844827586207
   'appointments' 0.860483393027
   'fedex' 0.895310019134
   'orders' 0.936008158023
   'url:main2' 0.949438202247
   'url:rx' 0.949438202247
   'url:php' 0.974186213206

The spammers are starting to squirm. :-)

  Neil



More information about the Spambayes mailing list