CEM

The official implementation for the paper CEM: Commonsense-aware Empathetic Response Generation.

Usage

Dependencies

Install the required libraries (Python 3.8.5 | CUDA 10.2)

pip install -r requirements.txt

Download Pretrained GloVe Embeddings and save it in /vectors.

Dataset

The preprocessed dataset is already provided as /data/ED/dataset_preproc. However, if you want to create the dataset yourself, delete this file, download the COMET checkpoint and place it in /data/ED/Comet. The preprocessed dataset would be generated after the training script.

Training

python main.py --model [model_name] [--woDiv] [--woEMO] [--woCOG] [--cuda]

Testing

For reproducibility, download the trained checkpoint, put it in a folder named saved and run the following:

python main.py --model cem --test --model_path save/CEM_19999_41.8034 [--cuda]

Evaluation

Create a folder results and move the obtained results.txt for each model to this folder. Rename the files to the name of the model and run the following:

python src/scripts/evaluate.py

Citation

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

@article{CEM2021,
      title={CEM: Commonsense-aware Empathetic Response Generation}, 
      author={Sahand Sabour, Chujie Zheng, Minlie Huang},
      journal={arXiv preprint arXiv:2109.05739},
      year={2021},
}

Sahandfer / CEM

CEM