SEARCH FOR: ‘Jon Smath’ and get => ‘John Smith’
via blog.eberly.org

Phonetic tokenization looks like a much better solution for correcting typos than fuzzy searching based on Levenstein distance.