Span Semantic Role Labeling on CoNLL 2005 (test)
87.7F1 ScoreLi et al. (b) AAAI (ELMo+biaffine)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Li et al. (b) AAAI (ELMo+biaffine)Syntax-aware=false, Sequence tagging=false, Architecture=ELMo+biaffine, Year=20192019.01 | 87.7 | — | — | |
| He et al. (a) (ELMo)Syntax-aware=false, Sequence tagging=false, Architecture=ELMo, Year=20182019.01 | 87.4 | — | — | |
| UniSRLEvaluation Setup=End-to-end, Model Variant=Single2019.01 | 86.3 | 85.2 | 87.5 | |
| He et al. (2018a)Evaluation Setup=End-to-end2019.01 | 86 | 84.8 | 87.2 | |
| Tan et al. (self-attention)Syntax-aware=false, Sequence tagging=false, Architecture=self-attention, Year=20182019.01 | 84.8 | — | — | |
| Strubell et al. (self-attention)Syntax-aware=false, Sequence tagging=true, Architecture=self-attention, Year=20182019.01 | 83.9 | — | — | |
| Strubell et al. (2018)Evaluation Setup=End-to-end2019.01 | 83.7 | 83.7 | 83.7 | |
| He et al. (highway BiLSTM)Syntax-aware=false, Sequence tagging=false, Architecture=highway BiLSTM, Year=20172019.01 | 83.1 | — | — | |
| Zhou and Xu (deep BiLSTM)Syntax-aware=false, Sequence tagging=false, Architecture=deep BiLSTM, Year=20152019.01 | 82.8 | — | — | |
| He et al. (2017)Evaluation Setup=End-to-end, Model Variant=Ensemble2019.01 | 82.7 | 82 | 83.4 | |
| He et al. (2017)Evaluation Setup=End-to-end, Model Variant=Single2019.01 | 81.2 | 80.2 | 82.3 | |
| Toutanova et al. (DP)Syntax-aware=true, Sequence tagging=false, Architecture=DP, Year=20082019.01 | 79.7 | — | — | |
| FitzGerald et al. (structured)Syntax-aware=true, Sequence tagging=false, Architecture=structured, Year=20152019.01 | 79.4 | — | — | |
| Punyakanok et al. (ILP)Syntax-aware=true, Sequence tagging=false, Architecture=ILP, Year=20082019.01 | 76.3 | — | — |