On Mar 17, 2005, at 9:37 AM, Michael Lewis wrote:

> I think SpamSieve is always in "training mode" since you can always 
> "mark
> as good" or "mark as spam" to tell SpamSieve how to classify something 
> it
> missed or got wrong.

Right. SpamSieve doesn't have separate modes; it learns and filters at 
the same time.

> The SpamSieve Manual says you should initially train your SpamSieve 
> with
> between 1000-2000 messages.

I'm glad you mentioned that, because it tells me that the sentence in 
the manual needs to be clarified. :-) The intended meaning is that you 
should initially train SpamSieve with a few hundred good and spam 
messages (if you have them around). Then, it will continue to learn 
from new messages that come in, and (most importantly) from its 
mistakes that you correct. The 1000-2000 is the size that the corpus 
will eventually grow to. You *could* train it with that many to start 
with, but it's not necessary, and most people don't have that much 
saved spam.

> By the way, I don't have "Use Entourage Address Book" checked because I
> use PowerMail! :)

:-) BTW, the "Use Mac OS X Address Book" option does apply to 
PowerMail, if you have PowerMail set to sync its address book.

-- 
Michael Tsai                                 <http://c-command.com>



Reply via email to