The Grumpy Editor's guide to bayesian spam filters
Posted Feb 26, 2006 15:09 UTC (Sun) by job
In reply to: The Grumpy Editor's guide to bayesian spam filters
Parent article: The Grumpy Editor's guide to bayesian spam filters
If you'd actually read Grahams paper you'd know that Bayesian auto classification of mail is a lot older than that. What Graham did was point out that many of the existing attempts had been "too smart" and that results could actually be more accurate when not converting character sets, cleaning up headers etc. since their very existence adds to the amount of classification information.
So while not being in any way groundbreaking, he pointed out some interesting points that really re-ignited the idea. That, and that is was probably a good time when many people was unable to manually cope with all spam, was what makes people refer to the paper so often.
to post comments)