Research map / Label-efficient and robust learning / Recognition and long-tail
Robust Named Entity Recognition with Truecasing Pretraining
Stephen Mayhew, Gupta Nitish, Dan Roth
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add pretrained truecaser representations to cased-trained BiLSTM-CRF+GloVe | BiLSTM-CRF+GloVe uncased · CoNLL2003 test, uncased | F1: 24.3 -> 84.5 | Improved |
| Add pretrained truecaser representations to cased-trained BiLSTM-CRF+GloVe | BiLSTM-CRF+GloVe uncased · Ontonotes test, uncased | F1: 83.5 -> 81.1 | Hurt |
| Add pretrained truecaser representations to cased-trained BiLSTM-CRF+GloVe | BiLSTM-CRF+GloVe uncased · WNUT17 test, uncased | F1: 15.7 -> 30.2 | Improved |
| Add pretrained truecaser representations to uncased-trained GloVe NER | BiLSTM-CRF+GloVe uncased · CoNLL test, uncased | F1: 87.3 -> 88.3 | Improved |
| Add pretrained truecaser representations to uncased-trained GloVe NER | BiLSTM-CRF+GloVe uncased · Ontonotes test, uncased | F1: 83.4 -> 84.4 | Improved |
| Add pretrained truecaser representations to uncased-trained GloVe NER | BiLSTM-CRF+GloVe uncased · WNUT17 test, uncased | F1: 41.1 -> 43.2 | Improved |
| Add pretrained truecaser representations to uncased BERT NER | BERT uncased · CoNLL test, uncased | F1: 91.0 -> 91.2 | Improved |
| Add pretrained truecaser representations to uncased BERT NER | BERT uncased · Ontonotes test, uncased | F1: 88.1 -> 88.1 | No clear effect |
| Add pretrained truecaser representations to uncased BERT NER | BERT uncased · WNUT17 test, uncased | F1: 46.1 -> 46.9 | Improved |
| Initialize the truecaser on CoNLL training data instead of Common Crawl (ablation) | BiLSTM-CRF+GloVe · CoNLL2003 testb, uncased | NER F1: 88.3 -> 85.3 | Hurt |
| Fine-tune a pretrained truecaser (ablation) | BERT uncased NER · WNUT17 test, uncased | NER F1: 46.9 -> 46.3 | Hurt |
| Fine-tune a pretrained truecaser (ablation) | BERT uncased NER · WNUT17 test, uncased | Char F1: 30.6 -> 52.3 | Improved |
Closest papers by meaning
- UniversalNER: Targeted Distillation from Large Language Models for Open Named Entity Recognition (ICLR 2024 · similarity 0.87)
- Span Graph Transformer for Document-Level Named Entity Recognition (AAAI 2024 · similarity 0.86)
- Adversity-aware Few-shot Named Entity Recognition via Augmentation Learning (AAAI 2025 · similarity 0.86)
- MCL-NER: Cross-Lingual Named Entity Recognition via Multi-View Contrastive Learning (AAAI 2024 · similarity 0.86)
- RoPDA: Robust Prompt-Based Data Augmentation for Low-Resource Named Entity Recognition (AAAI 2024 · similarity 0.85)
- PaDeLLM-NER: Parallel Decoding in Large Language Models for Named Entity Recognition (NeurIPS 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
