On Wed, Mar 10, 2004 at 05:30:36AM -0500, Andrew Schulman wrote: > > Yes, I read that. But I didn't find it very convincing. Unless there are > large errors in the implementation, one naive Bayesian filter is as good as > another. And SpamAssassin works well for me. > > Differences in resource use, I can believe.
Ah, but they're not the same. The major difference is that SpamProbe looks at pairs of words, not (just) single words. Implementations can also make other minor improvements which affect the hit rate. I too like SpamProbe but that word file can get big. 278MB on my primary platform. krb -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]