code for modular summarization work published in ACL2021 by Krishna et al

This repository contains the code for running modular summarization pipelines as described in the publication Krishna K, Khosla K, Bigham J, Lipton ZC. Generating SOAP Notes from Doctor-Patient Conversations." ACL 2021.

Although we can not release models trained on the confidential medical data, we have released models trained on the publicly available AMI dataset.

Download the ami_models folder from the link given below and put it at the root of the repository: https://drive.google.com/drive/folders/12Fkv_JvhJotvDTk2Z5PZYs4POIkqmwBZ?usp=sharing

To reproduce the results, you need to follow the following 3 steps.

# step1: downloads and preprocesses AMI dataset
./prepare_data.sh

 # step2: runs the summarization pipelines on the data and computes rouge scores
 # (before running this command, you need to download the models as shown above  )
./predict_ami.sh

# step3: print the results
python show_results.py