Assuming the OP just needs a name for lookup (rather than separate firstname and lastname), I would go the opposite direction:

my $normalized_name = lc $name;
$normalized_name =~ tr/a-z//cd;     # or =~ s/[^a-z]//g
[download]

All of the examples from the initial post would then become identical maggiesmiths. This would also work for the input MaggieSmith and Bubba-Joe Smith (producing bubbajoesmith) and for names which differ in cApiTaLiZatiOn.

But, like you said, the specifics are a little sparse, so this solution may also be unsuitable.