One additional point. While the Bayesian matcher is 'trained' with
specific examples, it applies the process by looking for similarities
rather than exact matches, so it generates a similarity score and that may
still result in 'mistakes' now and then. Generally, it gets better with
more
I suspect I corrupted the Bayesian reasoning, accidentally. I did an import and
matched things badly, then quit GC without saving. I understand there is a bug
report indicating that quit-before-save does not correct Bayesian errors.
So, how do you correct a Bayesian reasoning error that you