Awesome

COSPLAY

This is an early version code for the paper COSPLAY: Concept Set Guided Personalized Dialogue Generation Across Both Party Personas (SIGIR 2022) and a 10-minute video is provided for a quick and intuitive understanding of the paper.

Environment

We provide the conda env (cosplay.zip) to setup environment. With conda installed, create cosplay conda environment by:

unzip cosplay.zip
conda create --name cos --clone cosplay
conda activate cos

Data

The dataset and the preprocessed data for concept set framework can be found here (data.zip).

unzip data.zip

Train COSPLAY in Supervised Learning

python train_cosplay_in_supervised.py

Train COSPLAY in Reinforced Learning

python train_cosplay_in_reinforced.py

Evaluate COSPLAY

In addition to the above training from scratch, our pre-trained cosplay can be found here. Then test model by the evaluation codes.

Citation

If you use this codebase in your work, please consider citing our paper:

@inproceedings{xu2022cosplay,
  title={COSPLAY: Concept Set Guided Personalized Dialogue Generation Across Both Party Personas},
  author={Xu, Chen and Li, Piji and Wang, Wei and Yang, Haoran and Wang, Siyun and Xiao, Chuangbai},
  booktitle={Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval},
  pages={201--211},
  year={2022}
}