word chainer made for nanogenmo 2014. not a markov chain.

pseudomarkov.py

import random

class PseudoMarkov():
    def __init__(self):
        self.associations = {}

    # pass it the corpus filepath
    def gen(self, filepath):
        sentences = []

        with open(filepath) as fp:
            sentences = fp.readlines()

        for sentence in sentences:
            self.breakup(sentence.split())

        s = random.choice(sentences)
        words = random.choice(sentences).split()

        i = 0
        while random.randint(0, len(words)-1) > len(words)-1:
            i+=1

        seedword = words[i]
        prefix = " ".join(w for w in words[:i])
        text = prefix + " " + seedword + " " + self.chain(seedword)

        # don't know where the key errors are coming from, so just sweeping them
        # under the carpet with this
        while text.split()[-1] in self.associations:
            text += self.chain(text.split()[-1])

        return text.strip()

    # ughhh don't ask
    def breakup(self, sentence):
        max = len(sentence) - 1

        for i in range (0, max):
            if sentence[i] in self.associations:
                l = self.associations.get(sentence[i])
                if sentence[i+1] not in l:
                    if i >= max:
                        l.append(sentence[i].lower())
                    else:
                        l.append(sentence[i+1].lower())
            else:
                if i >= max:
                    self.associations[sentence[i]] = [sentence[i]]
                else:
                    self.associations[sentence[i]] = [sentence[i+1]]

        if len(sentence) > 1:
            self.breakup(sentence[1:])


    def chain(self, seedword):
        if seedword in self.associations:
            return random.choice(self.associations[seedword]) + " "
        else:
            return seedword + " "


if __name__ == "__main__":
    chainer = PseudoMarkov()
    print chainer.gen("corpus.txt")