We are evaluating a bunch of options at the moment for Fuzzy searches. One of our requirements is that mistyped words be matched. An OCR of CRA1G (with the number 1) must ma
This might not help you, but we've had this problem for names. In german, there are alot of similar sounding names: Bayer, Beyer, Baier would all sound the same. For this, we've created a seperate column using phonetic spelling. You might create a similar alphabet for your fuzziness, where "I" equals "1", and get similar results.
SQL Server 2008 supports thesauruses should get you to where you need. You would define your misspellings as synonyms. See http://msdn.microsoft.com/en-us/library/ms142491.aspx.
Edit: To my understanding, you couldn't do this using a wildcard or regular expression approach. You'd have to define common replacement set patterns for your words that you anticipate will be most commonly mis-typed.