On Tue, 19 Aug 2003 17:12, Nick Rout wrote: > there is no pint at all in training a bayesian filter on someone elses > mail. it says so in the docs for every such filter i have investigated. > even esr (bogofilter) says so, so it must be right :-)
I think they mean for the _non-spam_ corpus. That needs to be your own. For the _spam_ corpus you could use any bag of v14gr4 / nigeria / suzy-waiting-to-meet-you / make $$ on e-bay spams. Yuri
