Research map / Label-efficient and robust learning / Recognition and long-tail
A Label Dependence-Aware Sequence Generation Model for Multi-Level Implicit Discourse Relation Recognition
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 |
|---|---|---|---|
| Use LDSGM instead of the strongest listed baseline | PDTB IDRR · PDTB test, top-level | F1: 63.39 -> 63.73 | Helped |
| Use LDSGM instead of the strongest listed baseline | PDTB IDRR · PDTB test, second-level | F1: 38.28 -> 40.49 | Helped |
| Use LDSGM instead of the strongest listed baseline | PDTB IDRR · PDTB test, connective | F1: 10.45 -> 10.68 | Helped |
| Add GCN label embeddings (ablation) | LDSGM · PDTB test, second-level | F1: 39.32 -> 40.49 | Helped |
| Add level-specific label attention (ablation) | LDSGM · PDTB test, second-level | F1: 39.15 -> 40.49 | Helped |
| Use previous-level predictions to guide current predictions (ablation) | LDSGM · PDTB test, second-level | F1: 39.27 -> 40.49 | Helped |
| Add mutual learning training (ablation) | LDSGM · PDTB test, top-level | F1: 62.93 -> 63.73 | Helped |
| Add mutual learning training (ablation) | LDSGM · PDTB validation, connective | F1 | Mixed |
| Use label sequence generation instead of independent softmax classifiers (ablation) | LDSGM encoder-decoder · PDTB test, second-level | F1: 38.28 -> 40.49 | Helped |
| Use the top-down decoder instead of CRF decoding (ablation) | LDSGM encoder-decoder · PDTB test, second-level | F1: 38.82 -> 40.49 | Helped |
| Use mutual learning instead of ensembling decoder predictions (ablation) | LDSGM · PDTB test, connective | F1: 8.61 -> 10.68 | Helped |
Closest papers by meaning
- Improving Implicit Discourse Relation Recognition with Natural Language Explanations from LLMs (AAAI 2026 · similarity 0.90)
- Linguistic Properties Matter for Implicit Discourse Relation Recognition: Combining Semantic Interaction, Topic Continuity and Attribution (AAAI 2018 · similarity 0.90)
- Implicit Discourse Relation Classification via Multi-Task Neural Networks (AAAI 2016 · similarity 0.88)
- A Deep Sequential Model for Discourse Parsing on Multi-Party Dialogues (AAAI 2019 · similarity 0.87)
- Synergistic Anchored Contrastive Pre-training for Few-Shot Relation Extraction (AAAI 2024 · similarity 0.82)
- Momentum Pseudo-Labeling for Weakly Supervised Phrase Grounding (AAAI 2025 · similarity 0.82)
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.