Research map / Label-efficient and robust learning / Recognition and long-tail
Semi-Supervised Learning on Meta Structure: Multi-Task Tagging and Parsing in Low-Resource Scenarios
KyungTae Lim, Jay Yoon Lee, Jaime Carbonell, Thierry Poibeau
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 |
|---|---|---|---|
| replace concatenated-view baseline with META-BILSTM multi-view structure (ablation) | BASELINE parser · UD 2.3 low-resource 9-language avg | LAS: 36.2 -> 39.0 | Improved |
| add co-training consensus loss on unlabeled data (Co-meta ensemble) (ablation) | META-BASE · UD 2.3 low-resource 9-language avg | LAS: 39.0 -> 40.3 | Improved |
| ensemble-based vs voting-based target extraction (ablation) | Co-meta · UD 2.3 low-resource 9-language avg | LAS: 39.4 -> 40.3 | Improved |
| remove confidence score from consensus loss (ablation) | Co-meta · Greek el_bdt 100 train sents | LAS (META): 69.1 -> 67.4 | Hurt |
| add Co-meta SSL on Finnish with very few tokens (ablation) | META-BASE · Finnish fi_ftb | LAS: 21.9 -> 21.8 | No clear effect |
| out-of-domain unlabeled data (Wikipedia) instead of in-domain (ablation) | Co-meta Ensemble · Chinese zh_gsd | LAS: 45.3 -> 46.3 | Improved |
| out-of-domain unlabeled data (Wikipedia) instead of in-domain (ablation) | Co-meta Ensemble · Greek el_bdt | LAS: 69.0 -> 68.7 | Hurt |
| add Co-meta SSL in high-resource setting without LM (ablation) | META-BASE · English en_ewt full train | LAS: 86.95 -> 87.01 | No clear effect |
| add Co-meta SSL with ELMo (ablation) | META-BASE · English en_ewt full train | LAS: 88.28 -> 88.25 | No clear effect |
| add Co-meta SSL with BERT-Multi (ablation) | META-BASE · English en_ewt full train | LAS: 89.49 -> 89.52 | No clear effect |
| Co-meta with 5 labeled sentences (ablation) | META-BASE/BASELINE · Chinese zh_gsd | LAS | Mixed |
Closest papers by meaning
- Learn from Your Neighbor: Learning Multi-modal Mappings from Sparse Annotations (ICML 2018 · similarity 0.86)
- Robust Meta-Representation Learning via Global Label Inference and Classification (TPAMI 2023 · similarity 0.86)
- CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 2025 · similarity 0.85)
- Semi-Supervised Multi-View Multi-Label Learning with View-Specific Transformer and Enhanced Pseudo-Label (AAAI 2025 · similarity 0.85)
- Partial Multi-View Multi-Label Classification via Semantic Invariance Learning and Prototype Modeling (ICML 2024 · similarity 0.85)
- FRABench and UFEval: Unified Fine-grained Evaluation with Task and Aspect Generalization (ICLR 2026 · 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.
