Hannah I want to understand what semi-supervised learning means in machine learning. How does it combine labeled and unlabeled data during model training? Can someone also explain where this approach is commonly used in real-world applications?