trinary measures

[Greg Louis]

On 20040218 (Wed) at 0655:21 -0500, Tom Allison wrote:

> I am using the trinary measurement of Yes/No/Unsure with the default 
> cutoff of 0.10 and 0.95.
> 
> What experience do people have with using this as a means of deleting 
> Yes results from the mail server without looking at them.  Is this 99% 
> safe?  99.9999% safe?  80% safe?

Not safe, I would guess.  As others have said, you need to keep
checking until you have some personal feeling for how well bogofilter
is doing for you.  Of course bogotune can help with that once you've
accumulated a good-sized archive of messages to work with.

I use ternary classification too.  Right now I've got 0.27 and 0.5 as
my cutoff figures (recommended by bogotune), and they're working
beautifully for me (0.5% false negatives, fewer than 0.008% false
positives).  But it's extremely unlikely they'd work for everyone.  And
I still spend about a minute a day scanning the 500 or so spams
bogofilter catches for me every 24 hours, before I chuck them.

HTH...
-- 
| G r e g  L o u i s         | gpg public key: 0x400B1AA86D9E3E64 |
|  http://www.bgl.nu/~glouis |   (on my website or any keyserver) |
|  http://wecanstopspam.org in signatures helps fight junk email. |