<ul><li>Programmatic weak supervision (PWS) reduces human effort for labeling data by combining user-provided labeling functions (LFs) on unlabeled datapoints.</li><li>Quality of generated labels depends on the accuracy of the LFs, leading to the study of fixing LFs based on a small set of labeled examples.</li><li>Novel techniques are developed for repairing LFs by minimally changing their results on labeled examples to ensure evidence for correct labels and high accuracy of LFs.</li><li>LFs are modeled as conditional rules to enable selective output changes for inputs, improving LF quality based on small sets of labeled datapoints.</li></ul>

Refining Labeling Functions with Limited Labeled Data

Discover more