Patent US 7,207,004

A method includes identifying a document corpus, creating a dictionary based on words in the document corpus, identifying a potentially misspelled word in a document, and providing a word from the dictionary as likely to correspond to the potentially misspelled word.