Generating a script for a Moe's Tavern scene using a recurrent neural network with long short term memory (LSTM) cells. The dataset is a subset of this dataset, and includes scenes from Moe's Tavern from 27 seasons.
This is Project 3 of Udacity's Deep Learning Nanodegree Foundation program. The aim of this project is generate a new Simpsons's script based on existing script data using recurrent neural networks, and go through the process of building and tuning one in TensorFlow.
| Hyperparameter | Number |
|---|---|
| Epochs | 50 |
| Batch size | 128 |
| RNN size | 512 |
| LSTM layers | 1 |
| Sequence length | 25 |
| Learning rate | 0.005 |
Check out the output in a professionally stylized script format! (Click here for PDF)
