Training ham seems difficult

[Boris 'pi' Piwinger]

Andreas Pardeike wrote:

> I find it easy to feed "junk" IMAP mailboxes into bogofilter to
> train spam but what about ham? Most users either use POP or if
> they use IMAP then their imap clients moves the messages to their
> local machines or they delete the messages because there is no
> reason to save some messages after they have been read.

Certainly the best is to independently collect spam and ham,
this of course means you have to correct this collection.

> Also, I don't want to train all incoming messages because this
> seems to train spam as ham (or should I untrain the messages in
> the "junk" mailboxes before I train them as spam?

You are talking about false negatives. You only should
correct (untrain as ham and train as spam) those.

pi