Graph Enhanced Contrastive Learning for Radiology Findings Summarization

Code for Graph Enhanced Contrastive Learning for Radiology Findings Summarization

==========

This repo contains the PyTorch code following this code

Citations

If you use or extend our work, please cite our paper at ACL-2022.

@inproceedings{hu-etal-2022-graph,
    title = "Graph Enhanced Contrastive Learning for Radiology Findings Summarization",
    author = "Hu, Jinpeng  and
      Li, Zhuo  and
      Chen, Zhihong  and
      Li, Zhen  and
      Wan, Xiang  and
      Chang, Tsung-Hui",
    booktitle = "Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = may,
    year = "2022"
}

Requirements

• Python 3 (tested on 3.7)
• PyTorch (tested on 1.5)
• We run our experiments on three 32GB-V100

Data

We give an example about the data in the graph_construction/

Preparation

Remain to be origanized. Some of the code needs to be debug, plz use it carefully.

Graph Construction

We have given the example about the data format to construct the graph (each line is a radiology report). You might need to change the data path to you own data path.

cd graph_construction
python graph_construction.py

After finish graph construction. need to run sh precess_radiology.sh to further process data. For this step, you can obtain more information from the link (https://github.com/nlpyang/PreSumm). Note that you also need to change the 322-324 row in src/prepro/data_builder.py to your own data.

Training

change DATA_PATH to your data_path, To start training, run

sh train_model_abs_openi_CL.sh

Evaluation

change DATA_PATH Model_path to your data_path and model path and let the step to a specific number To start evaluation, run

sh test_openi.sh

Pre-trained model

you can download the pre-trained models from (the link passwd: co14).