Reproduce “CrossWeigh: Training Named Entity Tagger from Imperfect Annotations” 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

Zihan Wang, Jingbo Shang, Liyuan Liu, Lihao Lu, Jiacheng Liu, Jiawei Han

CrossWeigh identifies potential NER annotation mistakes by training cross-checking models without the same entity names, then reducing flagged sentences’ influence rather than changing their labels. Five experts corrected labels in 186 of 3,453 CoNLL03 test sentences. On the corrected test set, Flair’s entity-wise F1 was 94.18 (±0.06), versus 93.89 (±0.06) without CrossWeigh, averaged over five runs; disagreement does not prove error.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.