Make Your Own Bible Verses

I was reading this article on the unholy union of Garfield and Markov models and it reminded me that I'd never gotten around to trying out markov modelling first hand.

Over my lunch hour I cranked out the code at the bottom of this entry. And yes I write code over my lunch hour as a form of relaxation. What do you do?

In any case, the previous blog entry was the result of running this script on the book of Mark. Mark, Markov, get it?

Any way if you've ever wanted to generate your own biblical sounding text, here's your chance. Just download some scripture and crank up your python interpreter and away you go!

BTW, I'm really curious if Mark(ov) Chapter 1 get's any hits from people seeking Biblical inspiration. I know it inspired me.

import sys
import random
import itertools

#http://www.o-bible.org/download/kjv.txt

def make_frequencies(words):
    result = {}

    for i in range(2, len(words)):
        w1, w2, w3 = words[i-2:i+1]
        if (w1,w2) not in result:
            result[(w1,w2)] = {}
        if w3 not in result[(w1,w2)]:
            result[(w1,w2)][w3] = 0
        result[(w1,w2)][w3] += 1

    return result
    
def find_first_random(freqs):
    return get_random_ith_item(list(set([x[0] for x in freqs])))

def find_second_random(first, freqs):
    pair = get_random_ith_item(list(set(x for x in freqs if x[0] == first)))
    if not pair:
        return find_first_random(freqs)
    return pair[1]

def find_markov_random(first, second, freqs):
    markov_words = freqs.get((first,second), {})
    word = get_random_ith_item(itertools.chain(*[[x]*count for (x,count) in markov_words.items()]))
    if not word:
        return find_second_random(second, freqs)

    return word

def get_random_ith_item(stream):
    item = None
    for i,x in enumerate(stream):
        if random.randint(0,i) == 0:
            item = x

    return item
   
def markovize(freqs, output_length):
    result = [find_first_random(freqs)]
    result.append(find_second_random(result[0], freqs))
    output_length -= 2

    while output_length > 0:
        w1, w2 = result[-2:] 
        result.append(find_markov_random(w1,w2,freqs))
        output_length -= 1

    return result

def main():
    output_length = 500

    words = ' '.join(sys.stdin.readlines())

    word_frequencies = make_frequencies(words.split())
    print ' '.join(markovize(word_frequencies, output_length))

if __name__ == '__main__':
    main()