Supervised learning uses data that is completely labeled, whereas unsupervised learning uses no training data.
In the case of semi-supervised learning, the training data contains a small amount of labeled data and a large amount of unlabeled data.
