Marky is a simple library implementing markov-chains of any direction/stepping for text generation.
You can create a chain like so:
text = ...
chain = marky.chain(1, marky.word_tokenize(text))
If you have a problem with nltk leaving punctuation on the ends of words, you can use marky.fix_passage_punc to strip punctuation from the ends of words.
text = ...
chain = marky.chain(1, marky.fix_passage_punc(marky.word_tokenize(text)))
Here, 1 is the step. The step is how far ahead chain pairs words. For example, with a step of 1:
(a, b, c, d, e) would pair (a, b), (b, c), (c, d), (d, e)
Whereas a step of 2 would pair it to:
(a, c), (b, d), (c, e)
If the sign is negative, the pairing goes backwards. This is useful for generating from an end word (see the poetry example). A step of -1 would produce
(e, d), (d, c), (c, b), (b, a)
And -2 would produce
(e, c), (d, b), (c, a)
After you have the chain, you can use marky.take(chain, n) to take n generated words. This is based on the itertools.take recipe, as a MarkovChain instance is also an iterator. If you need to pass more options (check the code for them), you can use the full-featured get_word method instead of next.
The full simple example code is:
text = ...
chain = marky.chain(1, marky.fix_passage_punc(marky.word_tokenize(text)))
print ' '.join(take(chain, 100))
Note: word_tokenize requires nltk, a python library for NLP.