Mail::SpamAssassin::Plugin::Bayes - determine spammishness using a Bayesian classifier


This is a Bayesian-style probabilistic classifier, using an algorithm based on the one detailed in Paul Graham's A Plan For Spam paper at:

It also incorporates some other aspects taken from Graham Robinson's webpage on the subject at:

And the chi-square probability combiner as described here:

The results are incorporated into SpamAssassin as the BAYES_* rules.


bayes_stopword_languages lang (default: en)

Languages enabled in bayes stopwords processing, every language have a default stopwords regexp, tokens matching this regular expression will not be considered in bayes processing.

Custom regular expressions for additional languages can be defined in

Custom regular expressions can be specified by using the bayes_stopword_lang keyword like in the following example:

 bayes_stopword_languages en se
 bayes_stopword_en (?:you|me)
 bayes_stopword_se (?:du|mig)

Regexps are case-insensitive will be anchored automatically at beginning and end.

To disable stopwords usage, specify bayes_stopword_languages disable.

Only one bayes_stopword_languages or bayes_stopword_xx configuration line can be used. New configuration line will override the old one, for example the ones from SpamAssassin default ruleset (

bayes_max_token_length (default: 15)

Configure the maximum number of character a token could contain