Building large-scale registries from unstructured clinical notes using a low-resource natural language processing pipeline

Nazgol Tavabi, James D. Pruneski, Shahriar Golchin, Mallika Singh, Ryan Sanborn, Benton Heyworth, Assaf Landschaft, Amir A. Kimia, Ata M. Kiapour

Abstract: Highlights•A low-resource and interpretable classification approach (SE-K) can facilitate information extraction from clinical notes.•The SE-K approach had better performance than embedding-based approaches including state-of-the-art BERT models.•The SE-K approach was at least six times faster (on CPU) than embedding bas-ed models on GPU.•The proposed SE-K approach offers interpretability by listing the keywords used to make the classification by the model.