An Empirical Investigation into Deep and Shallow Rule Learning