Update: Sorry, I realized after posting this is not entirely clear. I'd also strip the punctuation and then do the split on spaces, I just would probably think of the transliteration operator before the substitute (because I wouldn't necessarily be thinking regex), and I'd probably stick with it, thinking it might be faster. Does anyone know if it is?
One possibility:
#! perl
use strict;
$_="Wilmer'?";
tr/a-zA-Z//csd;
print;