>> I work for a company that handles pharmaceuticals. You have *no idea*
>> how much fun it is trying to figure out spam filter to not filter out
>> legitimate business enquiries, yet still filter out spam... :-D
>
> I'm guessing a Bayesian classifier is almost useless in this case? :)

No, a Baysian classifier would probably work quite well. See, all the 
ham contains works like "viagra" in the same sentence as "double-blinded 
randomised trial", while all the spam contains "viagra" in the same 
sentence as "lowest price meds of guaranteed". ;-)

It's the keyword-based filters that fail horribly. (And not just because 
many of them don't let you edit the keywords...)

Post a reply to this message