The success of machine learning systems in real-world applications heavily relies on the quality of data annotations, data selection, and the annotations themselves. Data annotations account for at least 90% of the value and the ROI in real-world applications of machine learning systems.