[Spambayes] Clasification difficulties using pop3proxy / sb_server

Mark Howells Mark.Howells at softoption.com
Mon Dec 1 05:29:06 EST 2003

> -----Original Message-----
> From: Tony Meyer [mailto:tameyer at ihug.co.nz]
> > I have been getting a lot of 'unsures' (30% or so).  When I 
> > use the web interface to review the messages and look at the 
> > clues for each unsure, I see that the spam rating is often 
> > 0.5 or even higher - sometimes well within the Spam zone (I 
> > currently use 20% and 80% as my thresholds). So , why is the 
> > classifier rating them as unsure?
> To paraphrase Tim, because they don't resemble the messages 
> you have trained
> as ham more strongly than the messages you have trained as 
> spam (or vice
> versa).  

Ooops - I've just looked at my question again and realise that the question looks ridiculous ;)  What I _meant_ to write, was that I get some message that are classified as 'unsure' when the declared spam probability is 0.8 or higher.  i.e. the spam rating is higher than the spam threshold. 

> Clues!  Clues!  Give us the clues!

I'm not at home just now but will grab one when I get chance.

> > Maybe related to the above, I have noticed that the 'unsure' 
> > subject line is marked as such in the cache. 

> Yes, that's correct.  [Note that I incorrectly guessed otherwise when
> someone else asked pretty much this exact question last week. 

And I apologise for asking a repeat question :( I've been following the list for a while but somehow didn't read that thread - sorry.

I've lodged a bug report.



Outgoing mail is certified Virus Free.
Checked by AVG Anti-Virus (http://www.grisoft.com).
Version: 7.0.203 / Virus Database: 261.3.2 - Release Date: 11/27/2003

More information about the Spambayes mailing list