Research map / Large language models / Language, safety and interpretability
Improving Distantly Supervised Relation Extraction by Natural Language 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 NLI-based indirect supervision and consolidate distant labels with IPIN | DSRE-NLI · NYT2-Test | F1: 69.93 -> 77.66 | Helped |
| Use NPIN instead of IPIN, retaining NLI positives and intersecting negatives | DSRE-NLI · NYT1-Test | F1: 69.42 -> 75.98 | Helped |
| Enrich general relation templates with selected corpus-mined patterns (ablation) | NLI DeBERTa zero-shot RE · NYT1-(D+T) | F1: 61.63 -> 73.20 (+11.57) | Helped |
| Enrich general relation templates with selected corpus-mined patterns (ablation) | NLI DeBERTa zero-shot RE · NYT2-(D+T) | F1: 61.02 -> 70.63 (+9.61) | Helped |
| Add IPIN consolidation using general templates to distant-supervision training (ablation) | DSRE-NLI · NYT1-(D+T) | F1: 57.65 -> 70.50 (+12.85) | Helped |
| Add IPIN consolidation using general templates to distant-supervision training (ablation) | DSRE-NLI · NYT2-(D+T) | F1: 59.28 -> 74.58 (+15.30) | Helped |
| Add selected mined patterns to IPIN consolidation with general templates (ablation) | DSRE-NLI · NYT1-(D+T) | F1: 70.50 -> 74.67 (+4.17) | Helped |
| Add selected mined patterns to IPIN consolidation with general templates (ablation) | DSRE-NLI · NYT2-(D+T) | F1: 74.58 -> 75.92 (+1.34) | Helped |
| Replace entity masking with Typed Entity Marker representation (ablation) | BERT multi-class RE model · NYT1-Test | F1: 75.98 -> 69.19 | Hurt |
| Replace entity masking with Typed Entity Marker representation (ablation) | BERT multi-class RE model · NYT2-Test | F1: 75.68 -> 69.94 | Hurt |
| Fine-tune the NLI model on raw distant annotations instead of using it as a filter (ablation) | NLI DeBERTa · NYT2-(D+T) | F1: 61.02 -> 57.35 | Hurt |
| Fine-tune the NLI model on IPIN-consolidated annotations (ablation) | NLI DeBERTa · NYT2-(D+T) | F1: 61.02 -> 65.34 | Helped |
Closest papers by meaning
- Global Distant Supervision for Relation Extraction (AAAI 2016 · similarity 0.91)
- Distantly Supervised Entity Relation Extraction with Adapted Manual Annotations (AAAI 2019 · similarity 0.90)
- Reliable Data Generation and Selection for Low-Resource Relation Extraction (AAAI 2024 · similarity 0.89)
- Enhancing Low-Resource Relation Representations through Multi-View Decoupling (AAAI 2024 · similarity 0.87)
- Enhancing Relation Extraction via Supervised Rationale Verification and Feedback (AAAI 2025 · similarity 0.87)
- Well-Written Knowledge Graphs: Most Effective RDF Syntaxes for Triple Linearization in End-to-End Extraction of Relations from Texts (Student Abstract) (AAAI 2024 · similarity 0.86)
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.