crccheck / haikufinder Goto Github PK

Installation requires NLTK and its punkt package. First-time steps:

pip install nltk
python -m nltk.downloader punkt
python setup.py install

A script to begin processing is located in /scripts.

A Python hack to find "haikus" in English text. For the purposes of this module, a "haiku" is one or more complete sentences that, together, can be broken into groups of 5, 7, and 5 syllables. Each candidate haiku line, and then the entire haiku, has to make it through a few heuristics to filter out constructions that are likely to scan awkwardly (like verb phrases split across lines). Since this code doesn't really try to understand the texts, it might throw away a few legitimate phrases, and it certainly lets through some bad ones.

Any improvements would be welcomed.

The license, found in license.txt, is "modified BSD".

Example usage:

text = open("ulysses.txt", "r").read()
haikus = HaikuFinder(text).find_haikus()
for haiku in haikus:
    print haiku[0]
    print "    %s" % haiku[1]
    print haiku[2]
    print

If you find that the HaikuFinder doesn't seem to recognize some word you're using, or is counting its syllables incorrectly, you can say

HaikuFinder.add_word('shmeggegge', 3)
HaikuFinder.add_word('kvetch', 1)
haikus = HaikuFinder('''
	For this I should stay? To hear some shmeggegge kvetch about his lawsuit?
	''').find_haikus()

If you like, you can fork the project, modify the custom.dict, and submit a pull request. I'll accept all useful dictionary additions.