Python program. Enter a word, and it spits out all words whose phonemes are subset of the phonemes in the original word. It's like an anagram finder (one where not all the original letters have to be found in the anagram to count), except that it works phonically. Uses the CMU phonic dictionary, which is included (perhaps legally?).
Some result words are printed with *'s next to them, I don't remember what feature of the result that signifies, and the code is too cryptic for me to bother figuring out (I wrote this over 10 years ago). It may be that all the *'d results have the same emphasis levels on their respective phonemes that the original word does.