Would you like to see your presentation here, made available to a global audience of researchers?
Add your own presentation or have us affordably record your next conference.
Subjective NLP tasks like offensive language detection often suffer from annotator disagreement, leading to noisy labels. We propose Weak Ensemble Learning (WEL), a framework that models annotator disagreement by constructing and aggregating weak predictors from diverse annotator perspectives. WEL does not require annotator metadata and outperforms strong baselines across four benchmark datasets.
