Hey i never said it would be perfect, and I think last names should be considered here. It would still need human review,
bradgafelman, your query would not work because it fully depends on last names, AND sometimes there is an _ and sometimes a - and etc. Id' be happy to get a rough approximation and then send these 8000 emails to human review from there.
and yes it's quite possible that "John Doe" is really Susan McGillicuddy, but hey that's another thing to deal with..