On Mon, 2003-12-01 at 17:38, Arlo Gilbert wrote: > Is there a faster way to get SA working? Because 90 days is a long time > to ask somebody to wait.
You could train it with a publicly-available corpus of spam, but that sort of defeats the purpose of the bayesian filter. It's highly personalized. My mother-in-law, for example, loves to get stuff from QVC and other similar online shopping places. Those messages almost always score very high with "standard" scoring due to content. I certainly don't want that junk, but she does. One man's trash is another man's treasure. It's unfortunate, but that's how statistical analysis works. You have to have a large sampling before any results can be standardized (otherwise you get wildly skewed results), and since Bayes filters are so personal, you can really only train it with your own mail. -- Frank Pineau Hey, you know those Roman hackers? Man, were they I III III VII!
signature.asc
Description: This is a digitally signed message part
