This code was used for QA model development described in the paper titled "Question-Answering System Extracts Information on Injection Drug Use from Clinical Notes".
The src folder contains 2 sub-folders: modeling and data_prep
- The
modelingsub-folder contains codes for tokenizing QA samples, finetuning and testing the QA model - The
data_prepsub-folder contains codes for cleaning the notes, an example of note enrichment, an example of automated gold-standard answer extraction, an example of parsing rules, and an example of question-to-answer mapping.
The plot_data folder contains data tables used for the error plots in the paper.
The environment.yml file has all the necesary packages to create the Conda environment.