sa-learn - SpamAssassin

This is a discussion on sa-learn - SpamAssassin ; Dear all, should I run sa-learn on mails that already are classified as spam? -- Regards Lars Ebeling http://leopg9.no-ip.org Hobbithobbyist "I am not young enough to know everything." -- Oscar Wilde...

+ Reply to Thread
Results 1 to 3 of 3

Thread: sa-learn

  1. sa-learn

    Dear all,

    should I run sa-learn on mails that already are classified as spam?
    --
    Regards
    Lars Ebeling

    http://leopg9.no-ip.org
    Hobbithobbyist

    "I am not young enough to know everything."
    -- Oscar Wilde


  2. Re: sa-learn

    Lars Ebeling wrote:
    > Dear all,
    >
    > should I run sa-learn on mails that already are classified as spam?




    nobody knows

    - some people "train on error". this is what I do.
    - others "train on everything". the problem here is to get the
    "everything" to use for training.=, since you should only train on
    manually reviewed mail.


  3. Re: sa-learn

    Lars Ebeling wrote:
    > Dear all,
    >
    > should I run sa-learn on mails that already are classified as spam?

    Yes.. Even if they already score BAYES_99 and are classified as spam,
    there's still value in training them.

    Admittedly false negatives are more "urgent" to get trained, but there's
    value in training all spam.


+ Reply to Thread