Ok, I've got some workable sourcecode. It's rough around the edges right now, it assumes input files are named certain things and crashes if it can't find them, it doesn't accept command-line arguments, it spits out a lot of debug info, and it isn't commented. I'll update them periodicly. They require a list of words, seperated by endlines. For wordlists, I suggest the scowl list (
http://prdownloads.sourceforge.net/wordlist/scowl-6.zip). I combined a few of the English wordlists there with a special hacker wordlist, ommitting words that were two characters or shorter (wordlist2.zip) and another omitting words that were three characters or shorter (wordlist3.zip)