Building machine learning pipelines to extract structured data from unstructured text is a popular problem within an unpopular development lifecycle.