Re: Bayesian Filtering for Spam

in reply to Bayesian Filtering for Spam

I read, with great interest, Paul Graham's article on filtering for spam using a Bayesian scoring system of individual words found in spam vs. 'good' email.

The translation of Graham's LISP fragments is the subject of LISP translation help ??.

I did an implementation of Graham's algorithm yesterday, and will post it once I've cleaned it up and tuned it. It seems to work very well.

Comment on Re: Bayesian Filtering for Spam

Replies are listed 'Best First'.
Re: Re: Bayesian Filtering for Spam by Anonymous Monk on Aug 17, 2002 at 19:48 UTC
Wierd. I wrote one yesterday as well but found it didn't work very well at all unless the ratio of spam to legit messages in the training set were relatively close to 1:1. For a corpus I used just the 'bare' set in: http://www.iit.demokritos.gr/~ionandr/lingspam_public.tar.gz.	[reply]

In Section Meditations