Spelling word suggestion tool Copyright © 2008 Bart Massey ALL RIGHTS RESERVED This software is licensed under the "3-clause ('new') BSD License". Please see the file COPYING provided with this distribution for license terms. "thimk" (an old joke) is a command-line spelling word suggestion tool. You give it a possibly-misspelled word, and it spits out one or more properly-spelled words in order of likelihood of similarity. The idea and name for thimk came from an old program that used to hang around Reed College, probably written by Graham Ross and now apparently lost in the mists of time. See for the one very vague reference I've found on the web (in the SEE ALSO section of the referenced manpage). The current implementation is a bit more sophisticated than I recall the original being. By default it uses a prefilter that discards words with large edit distances from the target, then filters words with a different phonetic code than the target, then presents the top result sorted by edit distance. The Soundex and Phonix phonetic codes are designed for names, but seem to work about the same with other words. I follow the common practice of not truncating the codes for greater precision, although Phonix does truncate its final "sound" for greater recall. The latest change to the implementation is an addition of an optional precompiled SQlite database of phonetic codes for the entire dictionary, created with "thimk-makedb". This greatly speeds lookup, permitting reasonable performance on enormous dictionaries.