Denis Shaposhnikov wrote, on 16. mar 2007 14:06:
When we started to train dspam merged group, we checked messages with
"showFactors=on" and found that some headers are useless and lead to
classification errors.
Could you explain how to found that they are useless?
He already did, more or less. In dspam.conf (3.6.8/CVS and probably
earlier) there's a *global* dspam.conf setting 'Preference
"showFactors=on"'. If you activate this, you'll get a load of crud in
each message dspam header section showing what lead factors to each
judgment, together with the weighting - à la SpamAssassin. Try it, and
you'll see - it's virtually useless in production, but can help during
training.
Best,
--Tonni
--
Tony Earnshaw
Email: tonni at hetnet dot nl
