Spammers have found a way around the Bayesian filter and my word specific filters. They create a message and then convert it to a gif image. There is no "text" in the message at all so it gets through the filters.
My brother's solution is to filter at a higher level. He deletes entire blocks, such as those from the asian pacific region. I have done the same thing for the past day.
I look through the headers of the mail in my junk box and see if they have the same starting address. For instance the ones starting with 210., 211. and 220.. In order to be sure only the actual sending address is being used and not some other part of the message I include the preceding opening bracket, [220.
This web site can be used to find the owner of an address:
http://www.arin.net/whois/index.html
It looks like eliminating entire address blocks stops about 1/4 of my 100 spams per day, many of which would get past the other filters.
Scott