Stacked Generalization for Medical Concept Extraction from Clinical Notes

The goal of our research is to extract medical concepts from clinical notes containing patient information. Our research explores stacked generalization as a metalearning technique to exploit a diverse set of concept extraction models. First, we create multiple models for concept extraction using a variety of information extraction techniques, including knowledgebased, rule-based, and machine learning models. Next, we train a meta-classifier using stacked generalization with a feature set generated from the outputs of the individual classifiers. The meta-classifier learns to predict concepts based on information about the predictions of the component classifiers. Our results show that the stacked generalization learner performs better than the individual models and achieves state-of-the-art performance on the 2010 i2b2 data set.