I think this is dangerous territory you are venturing into. Updating a data-file on basis of another list of data which "sound" the same is a very loose definition to go by. Can you tighten the rules on when to consider a data record to be the same as an existing entry? Otherwise I'm afraid not even Perl will do a good job here.

The venerable Soundex algorithm is known to map wildly different words to the same "stem" and that may or may not be what you want.