The training data is annotated. Once the model is learned, it will be used to make predictions on testing data.