What is the algorithm class for identifying spelling errors without the use of a dictionary?

Question

Please consider the following situation: in a segment of text, there are five occurrences of the string "slnFile", and one occurrence of "snlFile". The latter is a misspelling, but note that the former word is not in a proper dictionary ("slnFile" is a variable name indicating a "visual studio solution file", making sense only to the author of the text segment).

I can think of a simple spell checking implementation myself: find all word pairs in text segment where spelling differs by one character, indicate any word with a frequency count of 1 as suspect. (I know this is not a perfect solution.)

My question: what is the name for the class of algorithms that deal with this problem?

Answer 1

Calculate the Damerau-Levenshtein distance between all words in the vocabulary. Flag those that occur very infrequently and have a particularly small distance to a word that occured frequently.