Research map / Probabilistic and scientific modeling / Molecules and proteins
Span-Based Semantic Role Labeling with Argument Pruning and Second-Order Inference
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 PAPN argument pruning (ablation) | First-order SRLN · CoNLL 2005 WSJ, gold predicates | F1: 83.43 -> 86.17 | Helped |
| Add PAPN argument pruning (ablation) | First-order SRLN · CoNLL 2005 Brown, gold predicates | F1: 72.22 -> 75.58 | Helped |
| Add PAPN argument pruning (ablation) | First-order SRLN · CoNLL 2005 WSJ and Brown, gold predicates | F1: 86.72 -> 88.00; 78.88 -> 80.23 | Helped |
| Train PAPN independently instead of jointly with SRLN (ablation) | PAPN and first-order SRLN · CoNLL 2005 WSJ and Brown, predicted predicates | F1: 86.15 -> 86.56; 76.84 -> 77.29 | Helped |
| Train PAPN independently instead of jointly with SRLN (ablation) | PAPN and first-order SRLN · CoNLL 2005 WSJ and Brown, gold predicates | F1: 87.78 -> 88.00; 79.83 -> 80.23 | Mixed |
| Add self-loop edge prediction (ablation) | First-order SRL model · CoNLL 2005 WSJ and Brown | F1: 86.48 -> 86.56; 77.18 -> 77.29; 84.34 -> 84.29; 71.79 -> 71.78 | Mixed |
| Add self-loop edge prediction (ablation) | Second-order SRL model · CoNLL 2005 WSJ and Brown | F1: 86.46 -> 86.61; 77.26 -> 77.45; 84.46 -> 84.5; 72.28 -> 72.7 | Helped |
| Use second-order inference instead of first-order inference (ablation) | Ours + RoBERTa · CoNLL 2005 WSJ and Brown, predicted predicates | F1: 86.57 -> 87.03; 79.44 -> 79.71 | Helped |
| Use second-order inference instead of first-order inference (ablation) | Ours + RoBERTa · CoNLL 2012, predicted predicates | F1: 84.17 -> 84.33 | Helped |
| Use second-order inference instead of first-order inference (ablation) | Ours + RoBERTa · CoNLL 2005 WSJ, Brown, and CoNLL 2012, gold predicates | F1: 88.17 -> 88.51; 81.92 -> 82.12; 87.12 -> 87.36 | Helped |
| Replace the MLP argument-span scorer with a biaffine scorer (ablation) | PAPN · CoNLL 2005 and CoNLL 2012 development sets | precision: 14.35 -> 20.52; 15.07 -> 22.33 | Helped |
| Increase PAPN pruning λ from 1.0 to 1.5 (ablation) | Biaffine PAPN · CoNLL 2005 and CoNLL 2012 development sets | recall: 99.77 -> 99.93; 99.52 -> 99.82 | Helped |
Closest papers by meaning
- Dependency or Span, End-to-End Uniform Semantic Role Labeling (AAAI 2019 · similarity 0.91)
- Encoder-Decoder Based Unified Semantic Role Labeling with Label-Aware Syntax (AAAI 2021 · similarity 0.90)
- Frame Semantic Role Labeling Using Arbitrary-Order Conditional Random Fields (AAAI 2024 · similarity 0.90)
- MAPLE: Many-Shot Adaptive Pseudo-Labeling for In-Context Learning (ICML 2025 · similarity 0.84)
- Graph-KV: Breaking Sequence via Injecting Structural Biases into Large Language Models (NeurIPS 2025 · similarity 0.83)
- Short-Context Dominance: How Much Local Context Natural Language Actually Needs? (NeurIPS 2026 · 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.