RuleCleaner repairs weak-supervision labeling functions by minimally changing their outputs on a few labeled examples using a MILP plus rule-tree refinement, improving global labeling accuracy on most tested datasets.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Refining Labeling Functions with Limited Labeled Data
RuleCleaner repairs weak-supervision labeling functions by minimally changing their outputs on a few labeled examples using a MILP plus rule-tree refinement, improving global labeling accuracy on most tested datasets.