Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Hierarchical Attention Network with Pairwise Loss for Chinese Zero Pronoun Resolution
Peiqin Lin, Meng Yang
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 |
|---|---|---|---|
| add self-attention layer (ablation) | HAN-PL · OntoNotes 5.0 Chinese test, Overall | F-score (%): Table 3: 58.6 vs 60.2 | Improved |
| add npc-to-zp attention (ablation) | HAN-PL · OntoNotes 5.0 Chinese test, Overall | F-score (%): Table 3: 59.4 vs 60.2 | Improved |
| add zp-to-npc attention (ablation) | HAN-PL · OntoNotes 5.0 Chinese test, Overall | F-score (%): Table 3: 58.2 vs 60.2; zp-to-npc more important | Improved |
| use pairwise-margin loss instead of softmax cross entropy (ablation) | HAN-PL · OntoNotes 5.0 Chinese test, Overall | F-score (%): performance falls sharply by 3.7% | Improved |
| add similarity constraint among correct antecedents (ablation) | HAN-PL · OntoNotes 5.0 Chinese test, Overall | F-score (%): Table 4: 58.6 vs 60.2 | Improved |
| HAN-PL vs best baseline Yin et al. (2018b), overall | HAN-PL · OntoNotes 5.0 Chinese test, Overall | F-score (%): outperforms the best baseline by 2.9% | Improved |
| HAN-PL vs best baseline on source MZ | HAN-PL · OntoNotes 5.0 Chinese test, MZ | F-score (%): 52.5 -> 52.6 | No clear effect |
| HAN-PL vs best baseline on source TC | HAN-PL · OntoNotes 5.0 Chinese test, TC | F-score (%): 53.2 -> 54.7 | Improved |
| HAN-PL vs Yin et al. (2018b) | HAN-PL · OntoNotes 5.0 Chinese test, BN | F-score (%): 5.6% (source BN) | Improved |
| larger candidate set size nc (harder resolution); HAN-PL gains over AttentionZP grow | HAN-PL · OntoNotes 5.0 candidate-set size | F-score (%) | Improved |
Closest papers by meaning
- Chinese NER with Height-Limited Constituent Parsing (AAAI 2019 · similarity 0.85)
- Translating Pro-Drop Languages With Reconstruction Models (AAAI 2018 · similarity 0.84)
- SENCR: A Span Enhanced Two-Stage Network with Counterfactual Rethinking for Chinese NER (AAAI 2024 · similarity 0.82)
- Chinese Spelling Correction as Rephrasing Language Model (AAAI 2024 · similarity 0.82)
- Localized Zeroth-Order Prompt Optimization (NeurIPS 2024 · similarity 0.82)
- Underspecification in Language Modeling Tasks: A Causality-Informed Study of Gendered Pronoun Resolution (AAAI 2024 · 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.
