Research map / Label-efficient and robust learning / Recognition and long-tail
Merging Weak and Active Supervision for Semantic Parsing
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 |
|---|---|---|---|
| Add 1,000 actively queried full MRs to weakly supervised training | NSM+MAPO on WikiSQL · WikiSQL test | execution accuracy: 75.6 -> 78.6 | Helped |
| Add 50 human-annotated MRs | cold-start NSM+MAPO · WikiTableQuestions test | execution accuracy: 8.5 -> 34.1 (+25.6) | Helped |
| Add 50 human-annotated MRs | warm-start NSM+MAPO · WikiTableQuestions test | execution accuracy: 42.7 -> 42.7(+0.0) | No clear effect |
| Select failed parses using correctness rather than random selection (ablation) | Wassp on WikiSQL · WikiSQL test, 1,000-query budget | execution accuracy: 76.4 -> 78.6 | Helped |
| Use uncertainty selection alone instead of correctness selection (ablation) | Wassp on WikiSQL · WikiSQL test, 1,000-query budget | execution accuracy: 78.6 -> 77.0 | Hurt |
| Combine uncertainty selection with correctness filtering (ablation) | Wassp on WikiSQL · WikiSQL test, 1,000-query budget | execution accuracy: 77.0 -> 79.0 | Helped |
| Select queries using failed-word coverage instead of correctness selection (ablation) | Wassp on WikiSQL · WikiSQL test, 1,000-query budget | execution accuracy: 78.6 -> 77.4 | Hurt |
| Select queries by clustering sentence embeddings instead of correctness selection (ablation) | Wassp on WikiSQL · WikiSQL test, 1,000-query budget | execution accuracy: 78.6 -> 77.5 | Hurt |
| Use MR sketches instead of fully specified MRs (ablation) | Wassp on WikiSQL · WikiSQL test, 1,000 queries | execution accuracy: 78.6 -> 77.2 (-1.4) | Hurt |
| Use MR sketches instead of fully specified MRs (ablation) | Wassp on WikiTableQuestions · WikiTableQuestions test, cold start, 100 queries | execution accuracy: 37.7 -> 35.3 (-2.4) | Hurt |
Closest papers by meaning
- Learning to Map Frequent Phrases to Sub-Structures of Meaning Representation for Neural Semantic Parsing (AAAI 2020 · similarity 0.90)
- Neural Semantic Parsing in Low-Resource Settings with Back-Translation and Meta-Learning (AAAI 2020 · similarity 0.89)
- Weakly Supervised Sequence Tagging from Noisy Rules (AAAI 2020 · similarity 0.86)
- BenchCLAMP: A Benchmark for Evaluating Language Models on Syntactic and Semantic Parsing (NeurIPS 2023 · similarity 0.86)
- Exemplar Guided Active Learning (NeurIPS 2020 · similarity 0.86)
- Evaluating the Zero-shot Robustness of Instruction-tuned Language Models (ICLR 2024 · similarity 0.84)
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.