Stable Style Transformer with Classifier for Text Style Transfer (INLG 2020)

The overall flow of our model

Requirements

1. Pytorch 1.2+
2. Python 3.5+
3. Huggingface Transformer
4. BERTScore

Basically, the code is supposed to run in the GPU environment.
If you do not have a GPU, it is recommended to modify the code and use it in a CPU environment. By default, the folder path is an absolute path, so please set the root according to your settings.

Datasets

1. Yelp and Amazon Dataset
2. Human reference-DRG
3. Human reference-DualRL

Train

Description based on the yelp dataset

cd generation_model/yelp

Step 1: Train classifier

cd classifier
python3 train.py

Step 2: Train generator

python3 train.py

Evaluation

Step 1: Finetune GPT

cd gpt2/yelp
python3 train.py

Step 2: Evaluate models with 4 metrics

cd evaluation/yelp/my_model/SST/

Check out generalization_eval_new.ipynb
Systems are evaluated using BLEU, classification accuracy, PPL, and BERTscore.

Citation

@inproceedings{lee-2020-stable,
    title = "Stable Style Transformer: Delete and Generate Approach with Encoder-Decoder for Text Style Transfer",
    author = "Lee, Joosung",
    booktitle = "Proceedings of the 13th International Conference on Natural Language Generation",
    month = dec,
    year = "2020",
    address = "Dublin, Ireland",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2020.inlg-1.25",
    pages = "195--204"
}