tttitles

A neural network that learns to generate interesting book titles.

Get Training Data

• Crawl: Run scrapy runspider title_scraper.py -o titles.json to get book titles from some listopia on Goodreads slike Best Book Titles.
• Clean: Run python data_preprocess.py titles.json titles.txt to get rid of the titles with annoying non-ASCII characters and save the clean version as plain text in titles.txt.

Training

• Run python train_and_save.py to train a char-RNN model and save it in *.pt format.
• Tune the parameters like learning_rate, dropout, etc.

Generate some fun book titles

• Run python sample.py to read the model from file and generate random fun book titles.