Research map / Graph learning / Graph self-supervised learning
Reinforcement Learning Based Graph-to-Sequence Model for Natural Question Generation
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 syntax-based static graph construction rather than dynamic graph construction | G2S+BERT+RL · SQuAD split-2 test | BLEU-4: 18.06 -> 18.30 | Helped |
| Add RL fine-tuning (ablation) | G2S_sta+BERT · SQuAD split-2 test | BLEU-4: 18.02 -> 18.30 | Helped |
| Add RL fine-tuning (ablation) | G2S_dyn+BERT · SQuAD split-2 test | BLEU-4: 17.56 -> 18.06 | Helped |
| Fine-tune BERT rather than keep it fixed (ablation) | G2S_sta+BERT+RL · SQuAD split-2 test | BLEU-4: 18.20 -> 18.30 | Helped |
| Add DAN for answer alignment (ablation) | G2S_sta · SQuAD split-2 test | BLEU-4: 12.62 -> 16.96 | Helped |
| Add DAN for answer alignment (ablation) | G2S_dyn · SQuAD split-2 test | BLEU-4: 12.58 -> 16.81 | Helped |
| Use both word-level and contextual-level DAN alignment instead of either alone (ablation) | G2S_sta · SQuAD split-2 test | BLEU-4: 15.92 -> 18.02; 16.07 -> 18.02 | Helped |
| Use BiGGNN rather than Seq2Seq encoding (ablation) | G2S_sta · SQuAD split-2 test | BLEU-4: 16.14 -> 16.96 | Helped |
| Use BiGGNN rather than GCN encoding (ablation) | G2S_sta · SQuAD split-2 test | BLEU-4: 14.47 -> 16.96 | Helped |
| Pass messages in both directions rather than forward-only (ablation) | G2S_sta · SQuAD split-2 test | BLEU-4: 16.53 -> 16.96 | Helped |
| Pass messages in both directions rather than backward-only (ablation) | G2S_sta · SQuAD split-2 test | BLEU-4: 16.75 -> 16.96 | Helped |
| Add linguistic features (ablation) | G2S_sta · SQuAD split-2 test | BLEU-4: 16.65 -> 18.30 | Helped |
Closest papers by meaning
- EQG-RACE: Examination-Type Question Generation (AAAI 2021 · similarity 0.89)
- Entity Guided Question Generation with Contextual Structure and Sequence Information Capturing (AAAI 2021 · similarity 0.88)
- Ask the Right Questions: Active Question Reformulation with Reinforcement Learning (ICLR 2018 · similarity 0.87)
- Enhancing Question Generation through Diversity-Seeking Reinforcement Learning with Bilevel Policy Decomposition (AAAI 2025 · similarity 0.87)
- Question Calibration and Multi-Hop Modeling for Temporal Question Answering (AAAI 2024 · similarity 0.86)
- The Consensus Game: Language Model Generation via Equilibrium Search (ICLR 2024 · 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.