Use Unicode normalization?

Unicode defines normalized forms for characters and character classes.

It might work to normalize strings to NFKD and remove any characters of class Mn (Nonspacing_Mark) (see [table 12](http://www.unicode.org/reports/tr44/#GC_Values_Table))

It might be necessary to specially handle conversions like ß to ss

See also [python stack overflow answer](http://stackoverflow.com/a/1410365/2607762)