Grammatical Error Correction on BEA 2019 (dev)
63.4F0.5 ScoreGRECO voting+ESC
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GRECO voting+ESCVoting bias=True, Features=Edit scores from ESC2023.10 | 63.4 | — | — | |
| ESCModel type=System Combination Method2023.10 | 63.09 | — | — | |
| GRECO votingVoting bias=True2023.10 | 62.22 | — | — | |
| EditScorerModel type=System Combination Method2023.10 | 61.66 | — | — | |
| GRECOModel type=Quality Estimation Combination2023.10 | 60.74 | — | — | |
| MEMTModel type=System Combination Method2023.10 | 60.72 | — | — | |
| T5-LargeModel type=Base GEC System2023.10 | 56.21 | — | — | |
| GECTOR XLNetModel type=Base GEC System2023.10 | 55.62 | — | — | |
| VERNetModel type=Quality Estimation Combination2023.10 | 54.8 | — | — | |
| GECTOR ROBERTaModel type=Base GEC System2023.10 | 54.18 | — | — | |
| Riken&TohokuModel type=Base GEC System2023.10 | 53.95 | — | — | |
| Kakao&BrainModel type=Base GEC System2023.10 | 53.27 | — | — | |
| UEDIN-MSModel type=Base GEC System2023.10 | 53 | — | — | |
| SOMEModel type=Quality Estimation Combination2023.10 | 52.23 | — | — | |
| GPT-2Model type=Quality Estimation Combination2023.10 | 52 | — | — | |
| Transformer + BIFI (ours)training=unsupervised, critic=LM-Critic2021.09 | 42.4 | 51.6 | 24.7 | |
| Transformer + BIFI with no critictraining=unsupervised, critic=none2021.09 | 34.5 | 43.8 | 18.7 | |
| Transformertraining=unsupervised, data=synthetic paired data as in Awasthi et al. (2019)2021.09 | 34.1 | 44.2 | 17.9 | |
| NeuQEModel type=Quality Estimation Combination2023.10 | 29.3 | — | — |