Research map / Label-efficient and robust learning / Recognition and long-tail
Weakly Supervised POS Taggers Perform Poorly on Truly Low-Resource Languages
Katharina Kann, Ophélie Lacroix, Anders Søgaard
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 |
|---|---|---|---|
| Use cross-lingual ambiguous-label annotation instead of unsupervised clustering | POS tagging on 15 truly low-resource languages · 15 truly low-resource languages (average) | accuracy: 0.3440 -> 0.4138 | Improved |
| Use cross-lingual most-frequent-tag annotation instead of ambiguous-label annotation | Cross-lingual POS tagging · 15 truly low-resource languages (average) | accuracy: 0.4138 -> 0.4106; difference between AMB and FREQ is 0.0032 | No clear effect |
| Add character-level word autoencoding auxiliary task to AMB (ablation) | AMB · 15 truly low-resource languages (average) | accuracy: 0.4138 -> 0.4225 | Improved |
| Add character-level word autoencoding auxiliary task to FREQ (ablation) | FREQ · 15 truly low-resource languages (average) | accuracy: 0.4106 -> 0.4117 | Hurt |
| Add a next-word log-frequency language-modeling auxiliary task (ablation) | PLA16 · 15 truly low-resource languages (average) | accuracy: 0.4138 -> 0.3899 | Hurt |
| Use higher-quality dictionaries and UD training texts instead of comparable resources | AMB · 5 high-resource languages (average) | accuracy: 0.43 -> 0.72 | Improved |
| Use higher-quality dictionaries and UD training texts instead of comparable resources | FREQ · 5 high-resource languages (average) | accuracy: 0.50 -> 0.77 | Improved |
Closest papers by meaning
- Weakly Supervised Sequence Tagging from Noisy Rules (AAAI 2020 · similarity 0.86)
- Stronger Than You Think: Benchmarking Weak Supervision on Realistic Tasks (NeurIPS 2024 · similarity 0.86)
- Merging Weak and Active Supervision for Semantic Parsing (AAAI 2020 · similarity 0.86)
- Breaking Physical and Linguistic Borders: Multilingual Federated Prompt Tuning for Low-Resource Languages (ICLR 2024 · similarity 0.83)
- Language Ranker: A Metric for Quantifying LLM Performance Across High and Low-Resource Languages (AAAI 2025 · similarity 0.83)
- Escaping Collapse: The Strength of Weak Data for Large Language Model Training (NeurIPS 2025 · similarity 0.82)
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.
