Quicksilver,
First, here is a list of modules that you may find helpful

• Text::Levenshtein
• String::Approx
• Text::Soundex
• Text::Metaphone
• Text::Brew
• Text::WagnerFischer

If you want to find the "word" that matched using some fuzzy approximation, you first have to define what a "word" is. Then all you do is parse your input text one "word" at a time, apply your measure, and report on match.

You will find it harder than you think to define a word though if you want anything more than just rudimentary /\b(\w+)\b/. This is the crux of the problem.