Dataset annotation is the process of labeling data to make it understandable and useful for machine learning models, often requiring human input to ensure accuracy and context relevance. It is crucial for supervised learning, as annotated datasets are used to train models to recognize patterns and make predictions based on new, unseen data.