Download corpus of #tweets for user and generate new tweets.

mockingbird.py

#!/usr/bin/env python
"""Download corpus of tweets for user SCREEN_NAME and generate new tweets.

"""
import os
import sys
import json
import argparse
import twitter
import markovify


parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument('--screen_name', help='twitter username', type=str)
parser.add_argument('--corpus', help='saved corpus json file', type=str)
args = parser.parse_args()

if not (args.screen_name or args.corpus):
    sys.exit(1)

if args.corpus:
    all_statuses = json.load(open(args.corpus, 'r'))
elif args.screen_name:
    api = twitter.Api(consumer_key=os.environ['CONSUMERKEY'],
                      consumer_secret=os.environ['CONSUMERSECRET'],
                      access_token_key=os.environ['ACCESSTOKENKEY'],
                      access_token_secret=os.environ['ACCESSTOKENSECRET'],
                      sleep_on_rate_limit=True)

    all_statuses = []
    older_than = None
    while True:
        statuses = api.GetUserTimeline(screen_name=args.screen_name,
                                       count=200,
                                       exclude_replies=True,
                                       include_rts=False,
                                       max_id=older_than)
        if not statuses: break
        older_than = statuses[-1].id - 1
        all_statuses.extend([x.text for x in statuses])

    json.dump(all_statuses, open(args.screen_name + '_corpus.json', 'w'))

# Build the model.
text_model = markovify.Text(all_statuses)

# Emit some sentences.
for _ in range(20):
    print text_model.make_short_sentence(140, tries=100)