Research map / Label-efficient and robust learning / Recognition and long-tail
Neural Character-level Dependency Parsing for Chinese
Haonan Li, Zhisong Zhang, Yuqi Ju, Hai Zhao
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 ROOT-CHAR+WORD character POS tags vs none (ablation) | Char LSTM Parser · CTB5 character-level, dev | UAS: 68.92 -> 89.51 | Improved |
| use ROOT-CHAR+WORD instead of CHAR+WORD tags (ablation) | Char LSTM Parser · CTB5 character-level, dev | LAS: 80.31 -> 87.45 | Improved |
| use ROOT-CHAR+WORD instead of CHAR+WORD tags (ablation) | Char LSTM Parser · CTB5 character-level, dev | UAS: 87.24 -> 89.51 | Improved |
| use word-level tags only instead of character-level tags only (ablation) | Char LSTM Parser · CTB5 character-level, dev | UAS: 71.87 -> 79.49 | Improved |
| add character position info to word tags (WORD+CHAR-POSIT vs WORD-ONLY) (ablation) | Char LSTM Parser · CTB5 character-level, dev | UAS: 79.49 -> 88.61 | Improved |
| use ROOT-CHAR+WORD instead of WORD+CHAR-POSIT (ablation) | Char LSTM Parser · CTB5 character-level, dev | LAS: 80.67 -> 87.45 | Improved |
| use LSTM parser instead of MaltParser at character level | Char parser · CTB5 character-level, test | UAS: 83.72 -> 90.00 | Improved |
| use gold instead of predicted character POS | Char LSTM Parser · CTB5 character-level, test | UAS: 82.53 -> 90.00 | Improved |
| character-level LSTM parser instead of word-level LSTM pipeline | parser · CTB5 pipeline CM | CM: 26.86 -> 34.36 | Improved |
| character-level parser with predicted POS vs word-level restoration comparison with Zhang 2014 | Our model · CTB5 word-level dependency | UAS: 82.14 -> 79.44 | Hurt |
Closest papers by meaning
- Cross-Lingual Universal Dependency Parsing Only From One Monolingual Treebank (TPAMI 2023 · similarity 0.87)
- A Neural Transition-Based Approach for Semantic Dependency Graph Parsing (AAAI 2018 · similarity 0.86)
- Attention-based Belief or Disbelief Feature Extraction for Dependency Parsing (AAAI 2018 · similarity 0.86)
- Deep Biaffine Attention for Neural Dependency Parsing (ICLR 2017 · similarity 0.85)
- Discontinuous Constituent Parsing with Pointer Networks (AAAI 2020 · similarity 0.85)
- Dependency Parsing is More Parameter-Efficient with Normalization (NeurIPS 2025 · 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.
