A lightweight, domain-adapted BERT model for clinical question answering, trained on a combination of SQuAD v2 and EMRQA-MSQuAD datasets.

Source code for the training script is available on GitHub. See eval_results.json for evaluation results, and train_results.json for training results.