News for hackers. Stuff that matters.
Re:Paul Graham provides answer to spam emails (2002) by danesparza on Friday May 03, 2019 @09:16AM (#19818099)
This article is 17 years old ... I must be missing something. Why is it relevant right now?
Re:Paul Graham provides answer to spam emails (2002) by megous on Friday May 03, 2019 @11:04AM (#19819050)
The technique still works and works excepionally well.

I only use a bayes filter (I use bogofilter) and out of some 15k spam e-mails and 150k ham e-mails I received in the last 9 months or so, I've seen like ~10 false negatives and around the same number of false positives. That's way better than gmail can ever dream of, where I always had issues with automated messages ending up in spam. With bogofilter, the training is always explicit, and otherwise the model doesn't change, so I can be sure that when my automated messages pass, they will always pass, and don't randomly stop passing "just because".

I started with a big archive (around 100k msgs each) of SPAM and HAM for the initial training. I learned from the early days of using e-mail to archive my spam, for the eventual training, since my first Linux job involved setting up a DSPAM installation.

Re:Paul Graham provides answer to spam emails (2002) by jwilk on Friday May 03, 2019 @11:08AM (#19819080)
HN has an obsession about Paul Graham.
Re:Paul Graham provides answer to spam emails (2002) by QuercusMax on Friday May 03, 2019 @02:48PM (#19821155)
I mean, this is literally Paul Graham's site.