Some day I may turn this into a chrome extension to highlight limericks on arbitrary pages. It would be joyous to see a highlighted section of some random wikipedia page pop out to be a limerick!
I made a bot that would fetch Twitter posts and re-tweet any it detected were 5-7-5 haiku. The syllable counter was just a series of regex replacements, loosely inspired by Metaphone, but it was about 95% accurate on my dictionary of 300,000 words.
I think the rhyme definition is wrong. To me a rhyme has to be have the same phonemes from the coda of the last accented syllable to the end of the line.
Yes, without accent detection, the 'limericks' it finds can generally only be read with forced stresses, and the rhyming schemes it winds up with can be quite weak (rhyming an unstressed 'the' at the end of two lines, for example).
Ironically, by not knowing how to properly handle a stress-unstress syllable pair at the end of a line, this program would completely miscue on any limericks it finds with a first line that ends in 'Nantucket'