Hi Bryan
> Given that auto_learn is working, is it still worthwhile to 
> run true-positives (spam) and true-negatives (hams) through 
> sa-learn, as the INSTALL document says you should do? 

Yes, it makes sense: auto-earn is quite conservative wrt to the messages it
actually trains from. This means that there's a lot of messages that, while
correctly identified as ham/spam, were still not used for autolearning.
Putting these through sa-learn after manual verification is certainly

Bye, Martin

Reply via email to