Training bayesian filters to recognize discussion topics in mailing list archives ?

We’ve discussed this idea at the OSS2008 WoPDaSD 2008 workshop with Sander Striker, and I’d be curious to know if it’s been attempted to analyse contents of mailing-lists with bayesian filter of spamassassin in order to detect particular topics of discussion, instead of filtering out spam.

Sometimes researchers on FLOSS projects try to analyse mailing-lists to detect communication patterns.

Maybe spamassassin could be used as a common bayesian analyser to be train to recognize subjects of discussions in order to help detect common topics in mailing-lists ? Of course this may have allready been tried… will be curious of our comments on that idea.

Update 2008/09/22 : more accurate title, as it is not all about spamassassin (see comment bellow).