過去のtweetsからマルコフ連鎖を用いたセンテンスを生成しtweet

getTweetsLog.py

#!/usr/lib/python
#vim:fileencoding=utf-8

import re
import twitter



def myApi(consumer_key, consumer_secret, access_token, access_token_secret):
    api = twitter.Api(consumer_key, consumer_secret, access_token, access_token_secret)
    return api 

consumer_key = 'hoge'
consumer_secret = 'hoge'
access_token = 'hoge'
access_token_secret = 'hoge'

api = myApi(consumer_key, consumer_secret, access_token, access_token_secret)


screen_name = 'screen_name'
filename = 'hoge.txt'

lists = api.GetUserTimeline(screen_name, count=200)

tweets = ''  
p = re.compile('\n|(@.+ )|(#.+)|(https?://[A-Za-z0-9\'~+\-=_.,/%\?!;:@#\*&\(\)]+)')

for s in lists:
    tweets += p.sub('', s.text)

idfile = open(filename, 'w')
idfile.writelines(tweets.encode('utf-8'))
idfile.close()

markovTweet.py

#!/usr/lib/python
#vim:fileencoding=utf-8

import random
import urllib
import urllib2
from BeautifulSoup import BeautifulSoup
import twitter

class markovTweet:
    def __init__(self, appid, consumer_key, consumer_secret, access_token, access_token_secret, filename):
        self.appid = appid
        self.api = twitter.Api(consumer_key, consumer_secret, access_token, access_token_secret)
        self.filename = filename

    def getTweets(self):
        tweets_file = open(self.filename, 'r')
        tweets = tweets_file.read()
        tweets_file.close()
        return tweets

    def yahooParse(self,tweets):
        pageurl = 'http://jlp.yahooapis.jp/MAService/V1/parse'
        results = 'ma'
        my_filter = '1|2|3|4|5|6|7|8|9|10|11|12|13'
        params = urllib.urlencode({'appid':self.appid, 'results':results, 'filter':my_filter, 'sentence':tweets})
        html = urllib2.urlopen(pageurl, params)
        return html

    def makeMarkov(self,wordlist):
        markov = {}
        w1 = (u'', u'')
        w2 = (u'', u'')
        for word in wordlist:
            if w1[0] and w2[0]:
                if(w1, w2) not in markov:
                    markov[(w1,w2)] = []
                markov[(w1,w2)].append((word[0], word[1]))
            w1, w2 = w2, word

        count1 = 0 
        count2 = 0 
        sentence = u"error:文頭に適切な品詞なし"
        while count1 < 50: 
            w1, w2 = random.choice(markov.keys())
            poslist = [u'接尾辞',u'動詞',u'助詞',u'助動詞',u'特殊']
            if w1[1] not in poslist:
                sentence = w1[0] + w2[0]
                cnt = random.randint(1,70)
                while count2 < cnt:
                    tmp = random.choice(markov[(w1,w2)])
                    sentence += tmp[0]
                    w1, w2 = w2, tmp
                    count2 += 1
                return sentence
                break
        return sentence

    def main(self):
        tweets = self.getTweets()
        html = self.yahooParse(tweets)

        soup = BeautifulSoup(html.read())
        wordlist = [(w.surface.string, w.pos.string) for w in soup.ma_result.word_list]

        sentence = self.makeMarkov(wordlist)
        self.api.PostUpdates(sentence.encode('utf-8'))

if __name__ == '__main__':
    appid = 'Yahoo!Japan アプリケーションID'
    consumer_key = 'hoge'
    consumer_secret = 'hoge'
    access_token = 'hoge'
    access_token_secret = 'hoge'
    filename = 'hoge.txt'

    obj = markovTweet(appid, consumer_key, consumer_secret, access_token, access_token_secret, filename)
    obj.main()