If you're sanitizing a database, and you know that the vast majority of words are capitalized correctly, then the problem is easy to solve. Go through the database, and for each name generate the lowercase version. Keep track of how many differently-cased forms correspond to the one common lowercase form ("mckenzie" vs "McKenzie" vs "Mckenzie"). The ones that rarely occur are the mistakes, the ones that often occur are correct.

You hope. :-)

Nat