The algorithm for allocation errors?

To calculate the Levenshtein distance between words, at each step the dynamics of remembering how he got there (the omission of letters, addition, replacement, matching the letter). Illumination to build the path, which obtained the minimum distance. If several vocabulary words have the same distance, choose from them the most common.

And you know in advance that the "jahn" — krivosheinym "john"? Or know, only that pair is a word from the dictionary? If the first — the easiest way stupid character for character.
If the latter — you can try a wildcard search. For example looking for the equivalent of "reccursion":
\r
r*n — too a lot of options?(+1 character)
re*n — still?
re*on... etc.
\r
accordingly, if you do not succeed, and error in the extreme letters — acting reverse(tecursion):
*ecursion -> *ecursio* -> *cursi* (for example — the correct version is found in the first iteration)
\r
This option usable SQL-LIKE tion and for a grep'