Controllable Mixed-Initiative Dialogue Generation through Prompting

Under construction! This repository primarily contains our code for RAP-GPT3, a chatbot built to function zero-shot in the PersuasionForGood context. However, you can also find the prompts used for the ESConv evaluation here.

ACL 2023 Citation:

@article{chen2023pre,
  title={Controllable Mixed-Initiative Dialogue Generation through Prompting},
  author={Chen, Maximillian and Yu, Xiao and Shi, Weiyan and Awasthi, Urvi and Yu, Zhou},
  journal={ACL 2023},
  year={2023}
}

paper link

The original RAP paper was presented at AACL 2022:

@inproceedings{chen2022seamlessly,
  title={Seamlessly Integrating Factual Information and Social Content with Persuasive Dialogue},
  author={Chen, Maximillian and Shi, Weiyan and Yan, Feifan and Hou, Ryan and Zhang, Jingwen and Sahay, Saurav and Yu, Zhou},
  booktitle={Proceedings of the 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th International Joint Conference on Natural Language Processing},
  pages={399--413},
  year={2022}
}

paper link

Original PersuasionForGood Dataset:

https://gitlab.com/ucdavisnlp/persuasionforgood/-/tree/master/data?ref_type=heads

Imitation Learning Classifier:

download here

labelencoder_A

labelencoder_B