Automatic query selection for acquisition and discovery of food-drug interactions
Résumé
Food-drug interactions can profoundly impact desired and adverse effects of drugs, with unexpected and often harmful consequences on the health and well-being of patients. A growing body of scientific publications report clinically relevant food-drug interactions, but conventional search strategies based on handcrafted queries and indexing terms suffer from low recall. In this paper, we introduce a novel task called food-drug interaction discovery that aims to automatically identify scientific publications that describe food-drug interactions from a database of biomedical literature. We make use of an expert curated corpus of food-drug interactions to analyse different methods for query selection and we propose a high-recall approach based on feature selection.