On 03/25/2011 07:22 AM, Kevin A. McGrail wrote: > On 3/25/2011 10:13 AM, Jari Fredriksson wrote: >> Yes, it contains "cialis". >> >> I just looked at my mass check results, and this came out. Dunno what >> could be done to it... >> > Agreed. The rule should have a low score or be used with other metas > because it clearly has known false positives. > > Some rules have to fire on ham. That doesn't make them bad rules.
However, it does make them sub-optimal. I've revised the rule from: header HK_NAME_DRUGS From:name =~ /(viagra|cialis)/mi to: header HK_NAME_DRUGS From:name =~ /(viagra|\bcialis|cialis\b)/mi My multi-lingual dictionary only has four matches for that now, judicialis, sternofacialis, viagram, and viagraph, all of which are limited to the american-english-insane dictionary obtained from http://wordlist.sourceforge.net/ (apt-get install wamerican-insane) and should be sufficiently rare as From names (or anything else for that matter). The previous pattern matched 287 words.
signature.asc
Description: OpenPGP digital signature
