Sentence Level Quality Estimation on BEA19 (test)
69.09PrecisionELECTRA-VERNet
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ELECTRA-VERNetbackbone=ELECTRA2021.05 | 69.09 | 60.91 | 67.28 | |
| BERT-GED (HYP)backbone=BERT, input=Hypothesis2021.05 | 68.18 | 53.85 | 64.73 | |
| BERT-VERNetbackbone=BERT2021.05 | 66.86 | 58.6 | 65.02 | |
| BERT-GED (JOINT)backbone=BERT, input=Joint Source + Hypothesis2021.05 | 66.8 | 55.09 | 64.07 | |
| BERT-QEbackbone=BERT2021.05 | 58.63 | 54.19 | 57.69 | |
| NQE (CR)variant=CR2021.05 | 57.92 | 47.43 | 55.47 | |
| NQE (RC)variant=RC2021.05 | 57.87 | 47.24 | 55.37 | |
| NQE (RR)variant=RR2021.05 | 57.22 | 46.33 | 54.65 | |
| NQE (CC)variant=CC2021.05 | 56.83 | 49.47 | 55.19 | |
| BERT-GED (SRC)backbone=BERT, input=Source only2021.05 | 47.15 | 65.09 | 49.9 | |
| BERT-LMbackbone=BERT2021.05 | 46.32 | 64.05 | 49.03 | |
| BERT-GQEbackbone=BERT2021.05 | 46.15 | 64.01 | 48.88 |