How do I train spamassassin?
Although the rules used by spamassassin are effective in catching a great amount of spam, the ability of spammers to get around those rules and the specific characteristics of mail received by a given user (and his/her preferences) may require some changes in those rules with time. In order to accomplish that, do the following: • Classify your mail into folders which only contain spam or non-spam (“ham”), but not both. For instance assume that you have saved spam messages in a folder called “junk”, and keep ham (good email) in a folder called “good_mail”. • Run sa-learn once per folder, telling sa-learn whether to consider that folders’ contents to be spam or ham: sa-learn –spam –mbox junk sa-learn –ham –mbox good_mail You may use more than one folder of each ham and spam. Just run sa-learn similarly for each folder. • Make sure that you train spamassassin with both ham and spam. If you only train with spam the filter will become biased towards spam. • You also want to clear all sp