Reproduce “A Scalable Framework for Automated NER Annotation Correction in Low-Resource Languages” 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
Toqeer Ehsan, Thamar Solorio
The authors improve training data for named entity recognition in Urdu, Shahmukhi, and Sindhi by standardizing word boundaries, adding missing-name labels, and iteratively proposing and filtering labels with multilingual models. Fine-tuning XLM-RoBERTa-large on corrected data raises MK-PUCIT test micro-F1 from 84.59 to 88.55; gains also occur on the other datasets, but added labels are not all verified.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.