Duplicate person Identification - typographical error function

We have a custom CRM named People360 with 650,000 names and over 800,000 addresses.  As new names enter our system , we have various duplicate person identification processes.  We have written a new one that looks at exact and fuzzy matches using weighted criteria that look at parts of a name, parts of addresses, email pieces and birthdates.  It is working rather well.  However, we want better. We would like to see a function that will find matches between objects where a typographical error has occurred, like 'Simoni' would be matched with 'Simini', or 'Farwell' be matched with 'Garwell'.  Both of the