ML models are highly dependent on massive sets of hand-labeled training data which are very expensive and time consuming to create — often requiring person-months or years to assemble, clean, and debug — especially when domain expertise is required.

OcularIT has focused on ways of programmatically generating training data—or, more succinctly, programming training data. We use Snorkel to write labeling functions (LFs) to automatically label unlabeled data to create training data for ML models. We write Python programs that take as input a data point from the input data and assign labels (or abstain) using heuristics, pattern matching, and third-party models.