Implementations of the models in the paper "Generating Informative and Diverse Conversational Responses via Adversarial Information Maximization" by Yizhe Zhang, Michel Galley, Jianfeng Gao, Zhe Gan, Xiujun Li, Chris Brockett, Bill Dolan, NeurIPS 2018

• Tensorflow (version >1.9)
• CUDA, cudnn
• pip install nltk gensim

• See ./run.sh for training and testing commands

• Not directly sharable due to copyright issue.
• Data format and screening criteria will be released soon. Stay tuned.
• See demo.p for data format.

For any question or suggestions, feel free to contact yizhe.zhang@microsoft.com

• Arxiv link: https://arxiv.org/abs/1809.05972

@inproceedings{zhang2018generating,
  title={Generating informative and diverse conversational responses via adversarial information maximization},
  author={Zhang, Yizhe and Galley, Michel and Gao, Jianfeng and Gan, Zhe and Li, Xiujun and Brockett, Chris and Dolan, Bill},
  booktitle={NeurIPS},
  year={2018}
}