emilyalsentzer/Bio_Discharge_Summary_BERT

The Bio_Discharge_Summary_BERT model was trained on all discharge summaries from MIMIC III, a database containing electronic health records from ICU patients at the Beth Israel Hospital in Boston, MA . The models were trained using code from Google's BERT repository on a GeForce GTX TITAN X 12 GB GPU . Each note was first split into sections using a rules-based section splitter . Each section was split into sentences using SciSpacy (en core sci md tokenizer) Then each section was divided into sentences with a tokenizer . All notes from the NOTEEVENTS table were included (~880M words). The models trained on were trained for 150,,