Reproduce “CleanCoNLL: A Nearly Noise-Free Named Entity Recognition Dataset” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Susanna Rücker, Alan Akbik
CleanCoNLL revises CoNLL-03 labels for named-entity recognition by adding Wikipedia links and combining them with model-based cross-checks and human review. For example, it changes William Hill from person to organization. Among sampled FLERT prediction errors, annotation errors fell from 47% on original CoNLL-03 to 6% on CleanCoNLL; these figures describe reviewed errors, not corpus-wide error rates.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.