This is implementation for the paper "Reinforcement Learning Based Text Style Transfer without Parallel Training Corpus" accepted by NAACL 2019.

1.yelp data needs to be reprocessed into tsf and org formats

2. The version of TensorFlow needs to be 1.X

3. This repo uses a pre-trained model as a test model