In English, "wrrd" leads to "ward" and "word" as suggestions, in that order. Words that have the same distance to the word in the text should probably be sorted by their occurrence. That will need to depend on a corpus of course, or on word occurrence lists. Alternatively, one could have to top 1000 words or so and prefer those.
There is a frequency dictionary available for English (BNC frequency list). But we could also create one from Wikipedia (but we need to ignore all formating words).
The infrastructure (number of occurrences in the dict files) for this will be in LT 2.4.