Annotation consensus: implications for passage recommendation in scientific literature