My python program for training bogofilter to distinguish between spam and non spam. It relies on the mail being in Maildirs on the machine the program is run on. It looks only at mail that has been read since the last time it was run. It assumes that this mail is in the correct folder so it compares where bogofilter thinks the mail should be, based on the X-Bogosity header, with where it actually is. Mail that isn't where bogofilter thinks it should be is fed back through bogofilter in the training mode so that bogofilter can improve. Works pretty well.

The program without html markup is here