FUZZY LOGIC SPELLING CHECKER
Original Publication Date: 1995-Jun-30
Included in the Prior Art Database: 2004-Apr-07
Xerox Disclosure Journal
A computer based spelling checker employing fuzzy logic is proposed. Fuzzy sets and membership functions are defined for labels comprising: typo, sounds like, and looks like. Labels allow for a heuristic knowledge of how users misspell words and to focus on a search for correct spelling. As a result, a richer set of suggestions is enabled without a compromise in performance. A "similarity" output is also defined to rank the order of suggestions. The fuzzy logic spelling checker will out perform a conventional spelling checker in its ability to suggest words that are similar to an unfamiliar word. By introducing fuzzy sets to linguistic labels such as "sounds like" and -looks like", a very large domain of potential matching words can be searched quickly and intelligently. By defining multiple levels of a match, suggestions can be presented in order of similarity. Hence, the first suggestion made to the user may be a list of words that are deemed to be '"very close" by virtue of their membership in the input sets that define similarity. If no match is found, an additional search may be initiated among a larger set of word which are deemed to be "close". The fuzzy logic spelling checker utilizes a standard spelling dictionary from which the alternatives can be chosen. A fuzzy algorithm is used to determine which of the dictionary entries should be proposed as the correct spelling for the unfamiliar word.