Skip to main content

Table 6 Cases from the institutional classification tasks where the results of CNN and the rule-based NLP algorithm are different

From: A clinical text classification paradigm using weak supervision and deep representation

Task Case # Text Snippets Gold Standard Rule-based NLP CNN
Mayo Clinic Smoking Status Classification 1 …She is a taxi driver and she has never used tobaco products… Non-smoker X Y
2 …No smoking after age XXX… Smoker X Y
3 …Tobacco current use: No never used any... Non-smoker Y X
Proximal Femur (Hip) Fracture Classification 1 …Indications: femur fx…Cannulated screw fixation of the right femoral neck… Proximal Femur fracture X Y
2 … Pin fixation across the proximal left femoral neck… Proximal Femur fracture X Y
3 Exam: Sp Cerv*2vw Flex/Ext only Indications: Fx Vertebra Cervical Closed… Non-Proximal Femur fracture X Y
4 Exam: R Major Jnt Asp and/or Inj Indications: R hip inj/marc/steroid; fx femur neck nos closed, pain hip... Proximal Femur fracture Y X
  1. We use Y to indicate the correct extraction result and X otherwise