If you just want to fix typos then it seems to me that Damerau-Levenshtein distance would be the best approach - a DL distance of 1 from a dictionary word is a

If you just want to fix typos then it seems to me that Damerau-Levenshtein distance would be the best approach - a DL distance of 1 from a dictionary word is a high likelihood of a transposition or a missing or extra character.