Grammatical Error Correction on NLPCC word-level (test)
64.51PrecisionTraditional Voting
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Traditional VotingThreshold T=42023.05 | 64.51 | 15.35 | 39.32 | |
| Traditional VotingThreshold T=32023.05 | 58.39 | 21.55 | 43.52 | |
| seq2seq-1Strategy=Single Model2023.05 | 46.17 | 29.51 | 41.48 | |
| Traditional VotingThreshold T=22023.05 | 45.58 | 34.66 | 42.88 | |
| Average of 4Strategy=Single Model Ensemble2023.05 | 43.52 | 30.61 | 40.1 | |
| seq2seq-2Strategy=Single Model2023.05 | 43.4 | 31.29 | 40.28 | |
| seq2edit-1Strategy=Single Model2023.05 | 43.08 | 30.05 | 39.64 | |
| Sentence-level EnsemblePLM Backbone=MacBERT-base-Chinese2023.05 | 42.24 | 34.15 | 40.33 | |
| Sentence-level EnsemblePLM Backbone=GPT2-Chinese2023.05 | 41.94 | 36.13 | 40.63 | |
| seq2edit-2Strategy=Single Model2023.05 | 41.41 | 31.58 | 38.98 | |
| Sentence-level EnsemblePLM Backbone=BERT-base-Chinese2023.05 | 41.38 | 24.55 | 36.39 | |
| Edit-combination EnsemblePLM Backbone=GPT2-Chinese2023.05 | 40.5 | 36.44 | 39.62 | |
| Edit-combination EnsemblePLM Backbone=MacBERT-base-Chinese2023.05 | 40.11 | 33.62 | 38.62 | |
| Edit-level EnsemblePLM Backbone=MacBERT-base-Chinese2023.05 | 40.07 | 32.87 | 38.39 | |
| Edit-level EnsemblePLM Backbone=GPT2-Chinese2023.05 | 39.44 | 36.07 | 38.71 | |
| Edit-combination EnsemblePLM Backbone=BERT-base-Chinese2023.05 | 37.56 | 23.94 | 33.72 | |
| Edit-level EnsemblePLM Backbone=BERT-base-Chinese2023.05 | 36.69 | 23.24 | 32.89 |