[linux-elitists] (tmda) Re: Constraining Bogus challenges.

Michael Shields shields@msrl.com
Fri Oct 3 19:47:46 PDT 2003


In message <20031003201412.GO7680@vitelus.com>,
Aaron Lehmann <aaronl@vitelus.com> wrote:
> Something tells me the Bayes weights are way too low by default.

For some reason Bayes is used aggressively if you have network tests
enabled.  I work around this in ~/.spamassassin/user_prefs:

    # Set Bayes rules to be the same with network enabled as they are
    # with network disabled.  This generally means positive scores are
    # higher.
    score BAYES_00 0 0 -5.300 -5.300
    score BAYES_01 0 0 -5.400 -5.400
    score BAYES_10 0 0 -5.300 -5.300
    score BAYES_20 0 0 -4.701 -4.701
    score BAYES_30 0 0 -1.070 -1.070
    score BAYES_40 0.0   
    score BAYES_44 0.0   
    score BAYES_50 0.0   
    score BAYES_56 0.0   
    score BAYES_60 0 0 1.997 1.997
    score BAYES_70 0 0 2.593 2.593
    score BAYES_80 0 0 5.300 5.300
    score BAYES_90 0 0 4.027 4.027
    score BAYES_99 0 0 5.200 5.200

One trick I thought of, which turns out to be very effective, is to
sort the spam folder by score so that the messages most likely to be
false positives are at the top.

    (add-to-list 'nnmail-extra-headers 'X-Spam-Status)
    (defun gnus-article-sort-by-spam-status (h1 h2)
      "Sort articles by score from the X-Spam-Status: header."
      (< (string-to-number (gnus-replace-in-string
                            (gnus-extra-header 'X-Spam-Status h1)
                            ".*hits=" ""))
         (string-to-number (gnus-replace-in-string
                            (gnus-extra-header 'X-Spam-Status h2)
                            ".*hits=" ""))))

-- 
Shields.




More information about the linux-elitists mailing list