Feature Stacking for Sentence Classification in Evidence-Based Medicine
Marco Lui · 2012
We describe the feature sets and methodology that produced the winning entry to the ALTA 2012 Shared Task (sentence classification in evidence-based medicine). Our approach is based on a variety of feature sets, drawn from lexical and structural information at the sentence level, as well level. We introduce feature stacking, a metalearner to combine multiple feature sets, based on an approach similar to the wellknown stacking metalearner. Our system attains a ROC area-under-curve of 0.972 and 0.963 on two subsets of test data. 1