On Friday, August 23, 2002, 12:58:36 AM, Dierk Haasis <[EMAIL PROTECTED]> wrote:
DH> Hello DVM! DH> On Thursday, August 22, 2002 at 10:31:56 PM you wrote: >> But I don't really consider myself qualified to judge the value of this >> idea. I'm a veterinarian, not a mathematician. DH> The mathematics are sound, the sample size is more than sufficient, DH> particularly considering that most spam is virtually identical (chain DH> letters, free porn passwords, new business schemes etc.). DH> The principle followed has been discussed here before, the main DH> recommendation coming down to "filter all legitimate mail into DH> folders, what is left is to a high probability spam, but inspect it DH> before deleting". DH> I am quite sure that the false positive number is wrong, it is DH> principally not feasible to rule out false positives on the basis of DH> empiricism. But you can bring down the number to zero "for all DH> practical purposes", that is, one or two false positives in one year DH> isn't that bad. Even normal post looses a letter or two sometimes. The SPAMTOOLS mailing list has a new thread, "Bayes Filtering", discussing this topic in considerable detail. For those who are interested, I have posted the entire thread here: http://www.bellsouthpwp.net/g/o/gold3062/Bayes%20Filtering.txt Using The Bat! v1.61 on Windows 98 4.10 Build 2222 A -- Regards, Morton A. Goldberg, DVM Plaza Pet Clinic 511 W. Baddour Pkwy. Lebanon, TN 37087 (615) 444-4703 mailto:[EMAIL PROTECTED] ________________________________________________ Current version is 1.61 | "Using TBUDL" information: http://www.silverstones.com/thebat/TBUDLInfo.html