This repository contains the source code for DeepReads, a collection of machine learning models for generating book descriptions by genre.
- DeepReads was created with Python 3.6.5.
- Training requires textgenrnn and TensorFlow.
The dataset in this repository is comprised of 5 files containing 1000 book descriptions each, separated by genre. There is a file for Fantasy, Mystery, Philosophy, Romance, and Science Fiction. Each file contains a book description on each line.
For example, the Fantasy file contains the book description:
"Taran wanted to be a hero, and looking after a pig wasn't exactly heroic, even though Hen Wen was an oracular pig. But the day that Hen Wen vanished, Taran was led into an enchanting and perilous world. With his band of followers, he confronted the Horned King and his terrible Cauldron-Born. These were the forces of evil, and only Hen Wen knew the secret of keeping the kingdom of Prydain safe from them. But who would find her first?"
The motivation behind this application of NLG is to create models which can perhaps inspire people in the realm of creative writing. Many writers sometimes have trouble generating ideas for what they would like to writ eabout. Furthermore, writers sometimes practice their abilities using prompts. If a particular Fantasy author would like to practice writing Fantasy or is in need of ideas for their next novel, DeepReads could potentially be of help.
- Install textgenrnn and TensorFlow.
- Navigate to the directory training.
- Run any of the files corresponding to each genre.
- The training parameters can be customized according to the textgenrnn documentation.
- Install textgenrnn and TensorFlow.
- Navigate to the directory generation.
- Run any of the files corresponding to each genre.
- The variable training_path can be modified to point towards a new model or any of the existing ones in the models folder.
<to be added>
<to be added>
<to be added>
<to be added>
This project was created entirely by me for the class Social Network Mining at FSU.
Enjoy!