The official Keras implementation of ACL 2020 paper "Pre-train and Plug-in: Flexible Conditional Text Generation with Variational Auto-Encoders".

If you use the dataset or code in your research, please kindly cite our work:

@inproceedings{duan-etal-2020-pre,
    title = "Pre-train and Plug-in: Flexible Conditional Text Generation with Variational Auto-Encoders",
    author = "Duan, Yu  and
      Xu, Canwen  and
      Pei, Jiaxin  and
      Han, Jialong  and
      Li, Chenliang",
    booktitle = "Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics",
    month = jul,
    year = "2020",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/2020.acl-main.23",
    pages = "253--262",
}

We put the conditional data (extracted from the original Yelp and News Title dataset) in the conditional data directory. You may want to download the original dataset to train PretrainVAE.

First, train Pretrain VAE with pretrainVAE.py for each dataset. Then you may want to train task-specific classifiers with Classifier_for_evaluation.ipynb. You can then train and evaluate PluginVAE with PPVAE-single.ipynb (for single-condition generation) and PPVAE.ipynb (for the standard setting).

Code credit: Yu Duan