DFMed: Medical Dialogue Generation via Dual Flow Modeling

This is the code for ACL 2023 Findings paper: Medical Dialogue Generation via Dual Flow Modeling by Kaishuai Xu, Wenjun Hou, Yi Cheng, Jian Wang, and Wenjie Li.

DFMed is a novel medical dialogue generation framework, which models the transitions of medical entities and dialogue acts via step-by-step interweaving.

Requirements

Please create a new conda env and install the following pytorch version and main requirement packages (others are in the file).

conda create -n dual_flow python=3.8
conda activate dual_flow

pip install torch==1.10.1+cu113 torchvision==0.11.2+cu113 torchaudio==0.10.1 --extra-index-url https://download.pytorch.org/whl/cu113

• transformers==4.24.0
• nltk==3.4.1
• rouge==1.0.1
• setuptools==59.5.0
• numpy

Please note that the nltk version is important for calculating BLEU.

Data and Checkpoints

Download the data (including the CMeKG knowledge graph, MedDG dataset, and KaMed dataset) through:

sh download_data.sh

Download the fine-tuned generation checkpoints through:

sh download_checkpoints.sh

Download the fine-tuned dual flow learning checkpoints through:

• MedDG at Dropbox link
• KaMed at Dropbox link

Results

Download the results of dual flow learning and response generation through:

sh download_results.sh

Directory

The final directory is as follows:

└── DFMed
    ├── dual_flow
    ├── generation
    ├── data
    ├── images
    ├── results
    │   ├── df_results
    |   └── generation_results
    ├── download_checkpoints.sh
    ├── download_data.sh
    ├── download_results.sh
    └── requirements.txt

Implementations

1. Train the dual flow learning model.

cd dual_flow

# For the MedDG dataset
sh train_meddg.sh

# For the KaMed dataset
sh train_kamed.sh

2. Get act and entity predictions from the checkpoints of top performance.

sh get_prediction_topk.sh

3. Train the generation model.

cd generation

# For the MedDG dataset
sh train_meddg.sh

# For the KaMed dataset
sh train_kamed.sh

4. Inference. The df_results directory contains act and entity predictions of our training.

cd generation

# For the MedDG dataset
sh eval_meddg.sh

# For the KaMed dataset
sh eval_kamed.sh

5. Calculate metrics. We use the algorithm presented by the official code of the MedDG dataset.

python metrics.py --hp ./generate.txt --rf ./reference.txt

Cite

If you use our codes or your research is related to our work, please kindly cite our paper:

@inproceedings{xu-etal-2023-medical,
    title = "Medical Dialogue Generation via Dual Flow Modeling",
    author = "Xu, Kaishuai  and
      Hou, Wenjun  and
      Cheng, Yi  and
      Wang, Jian  and
      Li, Wenjie",
    booktitle = "Findings of the Association for Computational Linguistics: ACL 2023",
    year = "2023",
    address = "Toronto, Canada",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2023.findings-acl.423",
    doi = "10.18653/v1/2023.findings-acl.423",
    pages = "6771--6784",
}