What characterizes semi-supervised learning?

A. It uses a small labeled dataset together with a larger unlabeled dataset B. It uses only unlabeled data C. It uses no input features D. It requires every observation to have multiple labels