bogus bogotuning

David Relson relson at osagesoftware.com
Wed Jan 28 00:53:08 CET 2004


Hello Jason,

If you _really_, _really_ want to run bogotune without meeting the
minimum counts, here's a short HOW-TO:

1) Find the error message in the code. 
2) Disable it the exit() call right after it.

That's all it takes.  However it's not a good idea.  Here're a couple of
the caveats:

All the messages you use for tuning _must_ be messages that you haven't
used for training.  There must be some scoring variation among the ham
and among the spam that are used for tuning.  Using previously scored
messages makes it impossible for bogotune to judge whether one set of
parameters is better than another.

Bogotune can be run without a previously built token database and will
use a percentage of the test messages to build a token database.  It
uses a sliding scale (based on the message count) to determine how much
to put in the token database.  The scaling routines will (probably) give
weird results if the minimum message count isn't present.

Note:  if you circumvent bogotune's efforts to help you, then you give
up the right to complain if bogotune's results _don't_ help you.

Enjoy!

David




More information about the Bogofilter mailing list