To find related words, you'll probably want to have a look at String::Approx and/or cpan:://Text::Levenshtein. For the data structure I'd use a set of sets (or a HoH) containing lists of two elements.

The hardest part will be to determine the definition of a group of "similar" strings. I.e. to determine the cut-offs for the distance functions.

Just my 2 cents, -gjb-

Update: Text::WagnerFischer could be useful too.