[Spambayes] A Spam Filter Evaluation

[Tony Meyer]

> There is a recent study report on results of eleven variants of six  
> widely used open-source spam filters including SpamBayes.
>
> [...]
> I’ll leave it to those with more time on there hands than I to  
> decipher the specific details on SpamBayes performance.
>
> The report is at
>
> http://plg.uwaterloo.ca/~gvcormac/spamcormack06.pdf

Thanks for this.  This study was one where filter developers  
themselves took part - Gordon (et al) didn't select the filters and  
do the evaluations independently (apart from a couple of extras).

I looked after the SpamBayes submission (as well as managing one of  
the test corpora).  If you want easier to read information about how  
SpamBayes did, my notebook paper or conference paper would be a good  
place (IMO) to start.

Note that I didn't submit SpamBayes hoping to get comparatively good  
results (although it did surprisingly well); the unsure range, which  
is fundamental to SpamBayes, makes getting good results in a  
evaluation like this very difficult.  The purpose was more to compare  
different methods of using SpamBayes (there were four variants  
evaluated).

=Tony.Meyer

-- 
Please always include the list (spambayes at python.org) in your replies
(reply-all), and please don't send me personal mail about SpamBayes.
http://www.massey.ac.nz/~tameyer/writing/reply_all.html explains this.