Learning from Multiple Noisy Partial Labelers

This work enables users to create partial labelers that output subsets of possible class labels would greatly expand the expressivity of programmatic weak supervision.