chrispmaag/generate_tv_scripts_with_lstm

Preprocess text data and then use LSTM to generate a TV script

Generate TV Scripts with LSTM

Summary

In this project, we will generate our own Simpsons TV scripts using RNNs, specifically an LSTM in TensorFlow. The neural network we build will generate a new TV script for a scene at Moe's Tavern.

Data

We will be using part of the Simpsons dataset of scripts from 27 seasons. We will be using a subset of the original dataset. It consists of only the scenes in Moe's Tavern.

Preprocessing Functions

def create_lookup_tables(text):
    """
    Create lookup tables for vocabulary
    :param text: The text of tv scripts split into words
    :return: A tuple of dicts (vocab_to_int, int_to_vocab)
    """
    # Narrow text down to unique words
    vocab = sorted(set(text))
    vocab_to_int = {word: ii for ii, word in enumerate(vocab)}
    int_to_vocab = dict(enumerate(vocab))
    return (vocab_to_int, int_to_vocab)

Created two dictionaries to go from words to ids and from ids to words. Next created another dictionary to help with tokenizing punctuation.

Choose Word

def pick_word(probabilities, int_to_vocab):
    """
    Pick the next word in the generated text
    :param probabilities: Probabilites of the next word
    :param int_to_vocab: Dictionary of word ids as the keys and words as the values
    :return: String of the predicted word
    """
    # If I want to return only the highest probability word, I can use idx = np.argmax(probabilities)
    idx = np.random.choice(len(int_to_vocab), p=probabilities)
    word = int_to_vocab[idx]
    return word

We'll select the next word using the softmax probabilities for each word.

Results

We trained on less than a megabyte of text and now have a sample script! For better results, we could use the broader Simpsons dataset.