This repository contains codes for Knowledge Source Intergration (KSI) framework in the paper: Bai, T., Vucetic, S., Improving Medical Code Prediction from Clinical Text via Incorporating Online Knowledge Sources, The Web Conference (WWW'19), 2019. This projec is to reproduce the results presented in the paper, explore the methodologies and possibly provide some potential improvements. You will also find a picture of the table of results, which are our reproducibility results.
I used the following environment for the implementation:
- python==3.7.0
- torch==0.4.1
- numpy==1.15.1
- sklearn==0.19.2
Before running the program, you need to apply for MIMIC-III dataset and put two files "NOTEEVENTS.csv" and "DIAGNOSES_ICD.csv" under the same folder of the project.
Once you get these two files, run preprocessing scripts python preprocessing1.py, python preprocessing2.py, python preprocessing3.py in order.
After running three preprocessing files, you can run any of four models (python KSI_LSTM.py, python KSI_LSTMatt.py, python KSI_CNN.py, python KSI_CAML.py) to see how much improvement KSI framework brings to the specific model.