Research map / Label-efficient and robust learning / Recognition and long-tail
Relation Extraction Exploiting Full Dependency Forests
Lifeng Jin, Linfeng Song, Yue Zhang, Kun Xu, Wei-Yun Ma, Dong Yu
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 1-best tree (GRN) with full dependency forest (DDCNN) | relation extraction model · BioCreative VI CPR test | F1: 51.4 -> 53.1 | Improved |
| add parser fine-tuning on RE loss to pretrained forest model (ablation) | FOREST-DDCNN · BioCreative VI CPR test | F1: 53.1 -> 55.7 | Improved |
| use data-dependent CNN instead of regular CNN for forests (ablation) | FOREST-DDCNN · BioCreative VI CPR test | F1: 50.5 -> 53.1 | Improved |
| encode 1-best trees with DDCNN instead of GRN (ablation) | TREE-GRN · BioCreative VI CPR test | F1: 51.4 -> 50.3 | Hurt |
| pretrain parser on treebank rather than random init with fine-tuning (ablation) | FORESTFT-DDCNN · BioCreative VI CPR test | F1: 45.4 -> 55.7 | Improved |
| use parser-derived forest probabilities instead of uniform weights (syntax) (ablation) | DDCNN · BioCreative VI CPR test | F1: 50.4 -> 53.1 | Improved |
| fine-tune randomly initialized parser on RE loss vs fixed uniform forest (ablation) | DDCNN · BioCreative VI CPR test | F1: 50.4 -> 45.4 | Hurt |
| FORESTFT-DDCNN vs TREE-GRN with only 1K parser training trees | relation extraction model · BioCreative VI CPR test | F1: 47.6 -> 51.5 | Improved |
| FORESTFT-DDCNN vs TREE-GRN with 5K parser training trees | relation extraction model · BioCreative VI CPR test | F1: 50.3 -> 53.7 | Improved |
| forest + parser fine-tuning vs 1-best tree | TREE-GRN · PGR test | F1: 78.9 -> 89.3 | Improved |
| forest + parser fine-tuning vs 1-best tree | TREE-GRN · SemEval-2010 Task 8 test | F1: 84.6 -> 85.5 | Improved |
Closest papers by meaning
- Multi-View Consistency for Relation Extraction via Mutual Information and Structure Prediction (AAAI 2020 · similarity 0.87)
- Global Greedy Dependency Parsing (AAAI 2020 · similarity 0.87)
- SEE: Syntax-Aware Entity Embedding for Neural Relation Extraction (AAAI 2018 · similarity 0.86)
- Dependency Parsing is More Parameter-Efficient with Normalization (NeurIPS 2025 · similarity 0.85)
- Enhancing Relation Extraction via Supervised Rationale Verification and Feedback (AAAI 2025 · similarity 0.84)
- Synergistic Anchored Contrastive Pre-training for Few-Shot Relation Extraction (AAAI 2024 · similarity 0.83)
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.
