> I cannot find any special setting for autolearn (either
> bayes_auto_learn_threshold_nonspam or
> bayes_auto_learn_threshold_spam ) in either
> /etc/mail/spamassassin/local.cf or
> ~/.spamassassin/user_prefs . It probably means they are the default
> which is
> 0.1 and 12.0 respectively which seems reasonable enough.
In your set, these are the scores of your ham-learnt messages:
1282213227.7828.1nhON:2,S 0.8
1282213227.7828.AHPct:2,S 0.8
1282213227.7828.CQy9k:2,S 0.8
1282213227.7828.VLBN5:2,S -2.0
1282213227.7828.zvzWO:2,S 0.8
Since they score 0.8, they wouldn't be learnt as ham with the default
bayes_auto_learn_threshold_nonspam value. So, probably at least this setting
had been overridden somewhere in your SA conf...
Also, since most of the above messages hit BAYES_50, I would say you have
network tests disabled in your setup.
If this is the case, I would suggest to either:
1) enable and appropriately configure network test, because they may help SA
to better classify your messages;
2) revise the default auto-learn threshold. Most of the SA user base
nowadays uses network test, which means that the default auto-learn
thresholds may reflect the wider spam-to-ham score spread this case yields.
Also note 1282213227.7828.VLBN5:2,S : is it really spam? Revise you ham
corpora: isn't there any spam?
> I'll try to retrain the Bayesian filter.
>
> Regards,
>
> Shlomi Fish
>
> --
> -----------------------------------------------------------------
> Shlomi Fish http://www.shlomifish.org/
> Understand what Open Source is - http://shlom.in/oss-fs
>
> God considered inflicting XSLT as the tenth plague of Egypt, but then
> decided against it because he thought it would be too evil.
>
> Please reply to list if it's a mailing list post -
> http://shlom.in/reply .
