Improving the learning process and providing more accurate similarity matrices for unannotated data can positively