Discontinuous constituent parsing on NEGRA (test)
92F1 ScoreF&R + BERTlarge
Evaluation results
| Method | Links | ||
|---|---|---|---|
| F&R + BERTlargeApproach Type=task-specific approaches, Pre-trained Model=BERT_LARGE2026.05 | 92 | 67.9 | |
| Corro + BERTxApproach Type=task-specific approaches, Pre-trained Model=BERT_X2026.05 | 91.6 | 66.1 | |
| F&R + dep + BERTbaseApproach Type=task-specific approaches, Pre-trained Model=BERT_BASE, Enhancement=dependency-informed2026.05 | 91 | 76.6 | |
| R&M + BERTbaseApproach Type=task-specific approaches, Pre-trained Model=BERT_BASE2026.05 | 90.9 | 72.6 | |
| in-order + SwapApproach Type=sequence-to-sequence models, Pre-trained Model=GottBERT_BASE, Linearization=in-order, Transition=Swap, Lexicalized=false2026.05 | 89.08 | 67.06 | |
| in-order + Swap#kApproach Type=sequence-to-sequence models, Pre-trained Model=GottBERT_BASE, Linearization=in-order, Transition=Swap#k, Lexicalized=false2026.05 | 88.93 | 65.38 | |
| top-down + SwapApproach Type=sequence-to-sequence models, Pre-trained Model=GottBERT_BASE, Linearization=top-down, Transition=Swap, Lexicalized=false2026.05 | 88.59 | 67.43 | |
| lexicalized in-order + Shift#KApproach Type=sequence-to-sequence models, Pre-trained Model=GottBERT_BASE, Linearization=in-order, Transition=Shift#K, Lexicalized=true2026.05 | 88.54 | 58.53 | |
| in-order + Shift#kApproach Type=sequence-to-sequence models, Pre-trained Model=GottBERT_BASE, Linearization=in-order, Transition=Shift#k, Lexicalized=false2026.05 | 86.76 | 46.86 | |
| lexicalized in-order + SwapApproach Type=sequence-to-sequence models, Pre-trained Model=GottBERT_BASE, Linearization=in-order, Transition=Swap, Lexicalized=true2026.05 | 86.75 | 57.71 | |
| F&RApproach Type=task-specific approaches2026.05 | 85.7 | 58.6 | |
| lexicalized top-down + SwapApproach Type=sequence-to-sequence models, Pre-trained Model=GottBERT_BASE, Linearization=top-down, Transition=Swap, Lexicalized=true2026.05 | 85.34 | 53.14 | |
| Stanojević and SteedmanApproach Type=task-specific approaches2026.05 | 83.6 | 50.7 | |
| Coavoux and CohenApproach Type=task-specific approaches2026.05 | 83.2 | 56.3 | |
| Coavoux et al.Approach Type=task-specific approaches2026.05 | 83.2 | 54.6 | |
| lexicalized in-order + Swap#KApproach Type=sequence-to-sequence models, Pre-trained Model=GottBERT_BASE, Linearization=in-order, Transition=Swap#K, Lexicalized=true2026.05 | 83.02 | 40.82 |