[Rspamd-Users] Autolearn rejected messages as spam

[J. Fahrner]

Am 2018-12-05 18:14, schrieb Ralph Seichter:
> I'm well aware of what you were aiming at. ;-) I am just wondering 
> about
> the possible benefit of training rspamd's bayes filter with data that
> was already classified as spam by other means?

Others mentioned it already: There are many reasons for a score >15, not 
only bayes results. To make the bayes filter better, it is an advantage 
to let it learn EVERY spam message, not only those it already has in its 
db.