I'm using DSPAM 3.6.2 on Redhat, running dspam in daemon mode and
doing some custom document classification using dspamc. I'm using the
hash_drv storage driver and configured to use markovian (algorithm =
naive, pvalue = markv, feature sbph).
Using a script similar to train.pl, I'm classifying a document, and
correcting DSPAM when it incorrectly classifies a doc as ham. I am not
using a delivery agent (hence --stdout).
Here's the classify command:
dspamc --classify --stdout --deliver=stdout --user durl < text
I'm capturing the result (X-DSPAM-Result). If a spam doc is classed as
innocent, I run something like this:
dspamc --process --stdout --user durl --signature=#{sig} --mode=tum
--class=spam --source=error < text
This works as expected. Subsequent tests of the spam are correctly
classified. Using wc on dspam_dump shows that additional tokens are
added to the database. However, dspam_stats is never updated... it
shows 0 for all data values.
I would have expected dspam_stats to increase the false negative count
in this situation.
Any ideas?
Adam
Received on Mon Nov 28 19:27:41 2005
This archive was generated by hypermail 2.1.8 : Tue Nov 29 2005 - 00:00:01 EST