On Thu, 29 Apr 2010 08:25:29 -0400
Frank Bures <lisfr...@chem.toronto.edu> wrote:
> I've been running spamassassin for years. I am using auto-learn with
> very conservative thresholds. However, after several years of usage
> my spam database is about three time larger than my ham database and
> I am starting to see false positives.
>
> Is there a way how to "shrink" the spam database?
Yes there is.
If you run sa-learn --backup you can process the database as a flat text
file e.g.
v 3 db_version # this must be the first line!!!
v 1716 num_spam
v 1281 num_nonspam
t 1 0 1244129543 152a127dd2
t 1 0 1244399507 f6796cae57
t 1 0 1244336958 d585b2c212
t 1 0 1244458917 ff5612c891
t 1 0 1244842267 1414bea872
...
s h d4303932c1106a6b39161c7c7166db0bd295a...@sa_generated
s h 09302920ab81b2442c9f95751d3f6aa3e76d9...@sa_generated
s s 43e88c7bc1f9c7645a07e044f6714044a6014...@sa_generated
s h 870687a6361e32521d0438c0a0ac494801caf...@sa_generated
where
v = metadata
t = tokens
s = signatures (for retraining)
and where the format for tokens is:
t Nspam Nham epoch-time hash
what you need to do write a script that divides the metadata num_spam
value and all the token Nspam counts by 3. The updated database can
then be loaded back in with --restore.
e.g.
awk '/^(t|v.*num_spam)/ { $2=int($2/3)} /^(v|s)/ || ($2+$3)>0
{print}' <oldfile > newfile
