[Spambayes] Training on unusual ham - any ideas?

Tim Peters tim.peters at gmail.com
Tue Jan 17 22:43:36 CET 2006


[bob posert]
>  Spambayes is great; I love it; can't say enough good things.  I was going
> crazy with > 100 spams per day and this has saved my sanity.  My family
> thanks you too - I'm in a much better mood now!

It is a lot better than suicide ;-)

> I wonder if you have any ideas on how to deal with one issue.  I subscribe
> to a book review email list.  The recommendations on the list are unlike any
> other mail I get.  I've been getting this for about a year and always have
> to recover it from spam - it's so unlike anything else I get that spambayes
> doesn't classify it the way I would like.
>
>  Notes: To make it even more difficult for SpamBayes, often sex is
> mentioned.  It's a grown-up review, but not an adult review.  You can see a
> sample at www.boldtype.com; then click on current.
>
>  Any ideas on how to get around this?

There is no way to know why a message scores as it does without
looking at the "spam clues".  Generate those for one of these
messages, and share them here.  There are many features in the sample
at www.boldtype.com that _should_ score as strong ham clues, provided
you've trained on previous email of this kind as ham, but it's not
possible to guess why those aren't working for you without seeing the
spam clues generated from your training database.


More information about the SpamBayes mailing list