Clean the database from non-spam mails?

[Chris Wilkes]

On Tue, Dec 02, 2003 at 07:14:10PM +0100, Johannes Klug wrote:
> 
> I'd like to remove all non-spam emails from my database. I
> trained bogofilter only with about 200 ham emails, now my ham-box
> is about 700.

I did a little experiment with removing all words whose ham counts were
higher than their spam counts, by filtering the output of bogoutil -d,
which is:
  1  word
  2  spam count
  3  ham  count
  4  date updated

  # bogoutil -d ./wordlist.db | awk '$2 > $3 {print}' | bogoutil -l ./new.db

However this causes all my emails that registered with a spamicity of
1.0000 to fall to 0.41.  Just looking for one obvious spam word
'enlarge' in an email:

  # bogoutil -d /tmp/orig/wordlist.db |  \
    awk '$2 > $3 {print}' | bogoutil -l /tmp/onlyspam/wordlist.db
  
  # bogoutil -w /tmp/orig/wordlist.db enlarge
                                 spam   good
                     enlarge       87      0

  # bogoutil -w /tmp/onlyspam/wordlist.db enlarge
                                 spam   good
                     enlarge       87      0

An email with 'enlarge' in it ($s = email message file):

  # bogofilter -vvv -d /tmp/orig/     -I $s | grep enlarge
    "enlarge"         87  0.000000  0.021712  0.999933 +

  # bogofilter -vvv -d /tmp/onlyspam/ -I $s | grep enlarge
    "enlarge"          0  0.000000  0.000000  0.415000 -

Doesn't seem to pick it up now.  Did I screw up something with creating
the new wordlist.db file?  The spam score of that email went from 1.0000
to 0.415000.  I'm running version 0.15.9.

Chris