Description: "we propose the task of learning interpretable relationships from open domain facts to enrich and refine concept graphs. The Bayesian network structures are learned from open domain facts as the interpretable relationships between relations of facts and concepts of entities."