Weight Aware Feature Enriched Biomedical Lexical Answer Type Prediction

Lexical Answer Type (LAT) prediction is an essential part of question classification. It aims to assign certain lexical answer type to the questions to narrow down the search space and improve the classifier’s performance. LAT prediction is a challenge in the biomedical domain since it is more of a multi-label classification question, which means each question has more than one label. In this paper, we employ the Label Powerset method to transform multi-label classification problems into multi-classification problems. Afterwards we introduced a random forest based mechanism to partition the features into used (important) and unused (unimportant) sets with corresponding weights. Furthermore, by assuming that the unimportant features are not useless, we employ principal components analysis to get the information from the unused feature set. By combing these two types of features, the experimental study on the BioMedLAT dataset has demonstrated our method’s potential.