An unlabeled example is one that has not been identified or categorized prior to presentation to the learner. The reason for using unlabeled examples is that they are often much less expensive to create or find since they can be generated automatically.
Unlabeled examples are interesting for learning because they force the learner to create a label and therefore may result in enhanced learning compared to labeled examples. This has also been described as the testing effect, which is the benefit to testing (labeling) as opposed to passive study.
A.Blum and T. Mitchell. Combining labeled and unlabeled data with co-training. In Proceedings of the 1998 Conference on Computational Learning Theory, July 1998.