Research map / Large language models / Language, safety and interpretability
Effective Modeling of Encoder-Decoder Architecture for Joint Entity and Relation Extraction
Tapas Nayak, Hwee Tou Ng
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 WDec instead of HRL | Joint entity and relation extraction · NYT29 | F1: 3.9% | Improved |
| Use PNDec instead of HRL | Joint entity and relation extraction · NYT24 | F1: 1.3% | Improved |
| Use a five-run ensemble instead of a single model | WDec and PNDec · NYT29 and NYT24 | F1: 4.2% and 3.5% higher F1 scores; 4.2% and 2.9% higher F1 scores | Improved |
| Add masking-based copy to WDec (ablation) | WDec · NYT29 and NYT24 | F1: 0.575 -> 0.626; 0.724 -> 0.775; 0.560 -> 0.610; 0.738 -> 0.776; 0.607 -> 0.682; 0.763 -> 0.817 | Improved |
| Use single attention rather than average or N-gram attention (ablation) | WDec · NYT29 and NYT24 | F1: 0.575, 0.560, 0.607; 0.724, 0.738, 0.763 | Improved |
| Use combo attention instead of decoder-hidden-only attention (ablation) | PNDec · NYT29 and NYT24 | F1: 0.663 -> 0.673; 0.785 -> 0.789 | Improved |
| Use combo attention instead of previous-tuple-only attention (ablation) | PNDec · NYT29 and NYT24 | F1: 0.665 -> 0.673; 0.784 -> 0.789 | Improved |
| Decode an entire tuple per step with pointers rather than one word at a time | PNDec compared with WDec · NYT29 and NYT24 | training and inference speed: more than two times faster | Improved |
| Decode an entire tuple per step with pointers rather than one word at a time | PNDec compared with WDec · NYT29 and NYT24 | GPU memory consumption: one-third | Improved |
Closest papers by meaning
- An Autoregressive Text-to-Graph Framework for Joint Entity and Relation Extraction (AAAI 2024 · similarity 0.89)
- Joint Extraction of Entities and Overlapping Relations Using Position-Attentive Sequence Labeling (AAAI 2019 · similarity 0.89)
- OneRel: Joint Entity and Relation Extraction with One Module in One Step (AAAI 2022 · similarity 0.88)
- Joint Entity and Relation Extraction with a Hybrid Transformer and Reinforcement Learning Based Model (AAAI 2020 · similarity 0.88)
- The Structure of Relation Decoding Linear Operators in Large Language Models (NeurIPS 2025 · similarity 0.86)
- DocKS-RAG: Optimizing Document-Level Relation Extraction through LLM-Enhanced Hybrid Prompt Tuning (ICML 2025 · 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.
