I have a business email and a personal email. Well, actually, I have dozens of business email addresses, and several personal ones, but I collect all this email together into two places, it's easier for me to handle it.
Until recently, I was running my despammer on the business account, but letting my personal account stay raw. However, the amount of spam that my personal account collected became so great, that I decided to run the same despammer on my personal account.
It's a fairly complex thing. It's written in perl, and some things are hard-coded, some things are in a database.
For example, if an email isn't sent to me (doesn't have my email address in the to: line), then it's spam. It an email was sent to several people, it's spam (or if it isn't actually spam, it's almost certainly so unimportant that I won't feel bad about not reading it). If the sender's server is "unknown" then it's spam. If it was sent by a mailed called "The Bat" then it's spam, and I have no idea what "The Bat" is, just that everything that it sends is spam. If it's mostly Chinese (or Japanese, or Russian) then if it isn't spam, it's still something I can't read, so into the bin it goes. If the word "unsubscribe" is there twice, it's spam. If a trojan or virus is attached, it's spam. How do I know if it's a trojan or virus? Simple - if the attachment is executable. Because no-one legitimate sends executables via email.
And so on. Some spam gets through, but not enough to be a nuisance, and if some spam does start arriving in volume, then I can easily add to the list of spammy phrases.
So, I decided to use the same despammer on my personal email, and to my great pleasure, it worked immediately. All the spam goes into one of the folders "non-roman", "executable", "spam" and "spam not sent to my email address". I give those a quite look and delete them.
I actually designed this in 1997, and (with some upgrades) it's worked well ever since. Maybe I should have marketed it. Of course, in 1997, hardly anyone used the internet, email wasn't common and spam almost unknown.