Inference and learning for directed probabilistic logic models door