Research map / Architectures and efficiency / Attention and transformers
Deterministic Attention for Sequence-to-Sequence Constituent 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 deterministic attention (datt-bound) to bottom-up linearization, single model | bottom-up seq2seq parser · WSJ test | F-score: 86.52 -> 88.57 | Helped |
| ensemble of five datt-bound models (ablation) | datt-bound · WSJ test | F-score: 88.57 -> 90.33 | Helped |
| add headword (Collins rules) to attention list (ablation) | datt-bound · WSJ test | F-score: 88.57 -> 88.22 | Hurt |
| combine datt-bound and datt-bound-head ensembles (ablation) | deterministic attention ensemble · WSJ test | F-score: 90.33 -> 90.60 | Helped |
| top-down instead of bottom-up linearization with probabilistic attention | probabilistic attention seq2seq · WSJ test | F-score: 86.52 -> 87.21 | Helped |
| deterministic attention (bottom-up) vs probabilistic attention (top-down), ensemble of five | seq2seq parser · WSJ test | F-score: 89.74 -> 90.33 | Helped |
| ensemble of five probabilistic top-down models | probabilistic top-down · WSJ test | F-score: 87.21 -> 89.74 | Helped |
Closest papers by meaning
- Improving Unsupervised Constituency Parsing via Maximizing Semantic Information (ICLR 2025 · similarity 0.86)
- Scaling Stick-Breaking Attention: An Efficient Implementation and In-depth Study (ICLR 2025 · similarity 0.85)
- Punctuation-aware Hybrid Trainable Sparse Attention for Large Language Models (NeurIPS 2026 · similarity 0.85)
- Short-Context Dominance: How Much Local Context Natural Language Actually Needs? (NeurIPS 2026 · similarity 0.84)
- Dependency Parsing is More Parameter-Efficient with Normalization (NeurIPS 2025 · similarity 0.84)
- Stack Attention: Improving the Ability of Transformers to Model Hierarchical Patterns (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.