Research map / Large language models / Retrieval and knowledge
How to Ask Better Questions? A Large-Scale Multi-Domain Dataset for Rewriting Ill-Formed Questions
Zewei Chu, Mingda Chen, Jing Chen, Miaosen Wang, Kevin Gimpel, Manaal Faruqui, Xiance Si
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 |
|---|---|---|---|
| train transformer on MQR instead of using GEC | question rewriting · MQR TEST | BLEU-4: 13.1 -> 22.1 | Improved |
| use transformer instead of LSTM seq2seq | MQR-trained rewriter · MQR TEST | BLEU-4: 19.2 -> 22.1 | Improved |
| add Quora duplicate pairs to training (ablation) | Transformer (MQR) · MQR TEST | BLEU-4: 22.1 -> 23.6 | Improved |
| add Paralex pairs to training (ablation) | Transformer (MQR) · MQR TEST | BLEU-4: 22.1 -> 21.7 | No clear effect |
| add Quora and Paralex pairs (ablation) | Transformer (MQR) · MQR TEST | BLEU-4: 22.1 -> 23.1 | Improved |
| add identity well-formed pairs (ablation) | Transformer (MQR) · MQR TEST | BLEU-4/ROUGE-L: BLEU-4 22.1 -> 21.1; ROUGE-L 56.6 -> 58.0 | Mixed |
| run GEC after transformer (ablation) | Transformer (MQR + Quora) · MQR TEST | BLEU-4: 23.6 -> 26.3 | Improved |
| run GEC before transformer (ablation) | Transformer (MQR + Quora) · MQR TEST | BLEU-4: 23.6 -> 24.8 | Improved |
| apply transformer twice (ablation) | Transformer (MQR) · MQR TEST | BLEU-4: 22.1 -> 20.4 | Hurt |
| add GEC after transformer (ablation) | Transformer (MQR + Quora) · MQR TEST human eval | grammar score: 0.75 -> 0.91 | Improved |
| use MQR-trained transformer instead of GEC | question rewriting · MQR TEST human eval | semantic equivalence wrt ill-formed: 1.00 -> 0.67 | Hurt |
Closest papers by meaning
- Putting People in LLMs’ Shoes: Generating Better Answers via Question Rewriter (AAAI 2025 · similarity 0.88)
- Rethinking Generative Large Language Model Evaluation for Semantic Comprehension (ICML 2024 · similarity 0.85)
- Is There No Such Thing as a Bad Question? H4R: HalluciBot for Ratiocination, Rewriting, Ranking, and Routing (AAAI 2025 · similarity 0.84)
- RepLiQA: A Question-Answering Dataset for Benchmarking LLMs on Unseen Reference Content (NeurIPS 2024 · similarity 0.84)
- SuRe: Summarizing Retrievals using Answer Candidates for Open-domain QA of LLMs (ICLR 2024 · similarity 0.83)
- TAIA: Large Language Models are Out-of-Distribution Data Learners (NeurIPS 2024 · similarity 0.83)
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.
