TwitterBot

Twitter bot that emulates a user

The TwitterBot class provides an easy method to connect to the twitter API via Tweepy and emulate a user. Methods used to model the language are

1. Markov
2. LSTM model
3. GRU model

The primary packages are the following (some of which are smaller projects, like Gramformer)

• Markovify
• Gramformer
• Tweepy
• Tensorflow
• Spacy

Prerequisites

1. Create a developer account with Twitter
2. Create and register an application
3. Generate the 4 unique keys
   • consumer_key
   • consumer_secret
   • access_token
   • access_token_secret
4. Create a file named 'keys.ini' (optional)

[twitter]
access_token = <access_token>
access_token_secret = <access_token_secret>
api_key = <api_key>
api_key_secret = <api_key_secret>

Installation

pip install git+https://github.com/zpreator/TwitterBot

python -m spacy download en_core_web_sm

Usage

Create a TwitterBot object using a username and keys.ini file. The keys in the file should be linked to the account which the bot will be tweeting from. The username passed into the constructor is the username of the tweeter you would like to emulate.

from twitter_bot import TwitterBot

if __name__ == '__main__':
    access_token = ""
    access_token_secret = ""
    api_key = ""
    api_key_secret = ""
    bot = TwitterBot('@laurarawra',
                     access_token=access_token,
                     access_token_secret=access_token_secret,
                     api_key=api_key,
                     api_key_secret=api_key_secret)
    bot.train_model(pull_tweets=True, model_type='markov')
    # bot.load_model(model_path='c:/path/to/model', model_type='markov')  # The model is stored when trained and can be loaded
    print('Generating a word: ', bot.generate_word())
    print('Generating a sentence: ', bot.generate_sentence())
    print('Reading tweets: ', bot.get_tweets())
    print('Tweeting message: ', bot.tweet_random_sentence())

Same thing, except this time using a .ini file to store the keys.

from twitter_bot import TwitterBot

if __name__ == '__main__':
    bot = TwitterBot('@<twitter-username>', keys_path='c:/path/to/keys.ini')
    bot.train_model(save_path='c:/path/to/model', pull_tweets=True, model_type='markov')  # This may take some time
    # bot.load_model(model_path='c:/path/to/model', model_type='markov')  # The model is stored when trained and can be loaded
    print('Generating a word: ', bot.generate_word())
    print('Generating a sentence: ', bot.generate_sentence())
    print('Reading tweets: ', bot.get_tweets())
    print('Tweeting message: ', bot.tweet_random_sentence())

Advanced Usage

This is an example of training an LSTM model with access to the model hyperparameters

from twitter_bot import TwitterBot

if __name__ == '__main__':
    bot = TwitterBot('@<twitter-username>')
    bot.train_model(pull_tweets=True,
                    model_type='lstm',
                    rnn_units=256,
                    batch_size=32,
                    num_epochs=10)
    print(bot.generate_sentence(num_chars=280, seed='start'))