Skip to content

Latest commit

 

History

History
69 lines (53 loc) · 2.95 KB

README.md

File metadata and controls

69 lines (53 loc) · 2.95 KB

Code for Supporter

The implementation of our paper accepted by ACL 2023: Facilitating Multi-turn Emotional Support Conversation with Positive Emotion Elicitation: A Reinforcement Learning Approach

venue status update

Requirements

  • Python==3.8.0
  • torch==1.10.0
  • transformers==4.1.1
  • Download Blender model 90M, and put it into the blender folder

Dataset

  • The preprocessed dataset is already provided at Google Driven. Change the folder name to data.

  • If you want to create the dataset yourself, download the comet-atomic-2020 (BART) checkpoint and place it in /data/ConstructDataset/Comet. The preprocessing details could be found in the main.sh script.

Rewards

Emotional Support Reward Model

Dialogue Coherence Reward Models

  • The trained Dialogue Coherence Reward Models is already provided at Google Driven.

  • Download bert-base-cased, and put it into the rewards/bert

  • If you want to train the Dialogue Coherence Reward Models yourself:

    cd rewards
    python construct_dataset.py
    cd ..
    bash main_rewards.sh
    

Training, Testing and Evaluating

bash main.py

Citation

If you find our work useful for your research, please kindly cite our paper as follows:

@inproceedings{DBLP:conf/acl/ZhouCWH23,
  author       = {Jinfeng Zhou and
                  Zhuang Chen and
                  Bo Wang and
                  Minlie Huang},
  editor       = {Anna Rogers and
                  Jordan L. Boyd{-}Graber and
                  Naoaki Okazaki},
  title        = {Facilitating Multi-turn Emotional Support Conversation with Positive
                  Emotion Elicitation: {A} Reinforcement Learning Approach},
  booktitle    = {Proceedings of the 61st Annual Meeting of the Association for Computational
                  Linguistics (Volume 1: Long Papers), {ACL} 2023, Toronto, Canada,
                  July 9-14, 2023},
  pages        = {1714--1729},
  publisher    = {Association for Computational Linguistics},
  year         = {2023},
  url          = {https://aclanthology.org/2023.acl-long.96},
  timestamp    = {Thu, 13 Jul 2023 16:47:40 +0200},
  biburl       = {https://dblp.org/rec/conf/acl/ZhouCWH23.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}