Sentence Level Quality Estimation on FCE (test)
58.77PrecisionELECTRA-VERNet
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ELECTRA-VERNetbackbone=ELECTRA2021.05 | 58.77 | 41.86 | 54.37 | 48.12 | |
| BERT-GED (JOINT)backbone=BERT, input=Joint Source + Hypothesis2021.05 | 58.53 | 37.24 | 52.53 | 45.08 | |
| BERT-VERNetbackbone=BERT2021.05 | 58.32 | 39.99 | 53.42 | 47.19 | |
| BERT-GED (HYP)backbone=BERT, input=Hypothesis2021.05 | 57.21 | 36.03 | 51.19 | 43.48 | |
| NQE (RC)variant=RC2021.05 | 53.97 | 31.35 | 47.17 | 31.2 | |
| BERT-QEbackbone=BERT2021.05 | 52.01 | 36.89 | 48.07 | 33.84 | |
| NQE (CR)variant=CR2021.05 | 51.77 | 31.46 | 45.85 | 30.69 | |
| NQE (RR)variant=RR2021.05 | 51.43 | 30.36 | 45.16 | 28.74 | |
| NQE (CC)variant=CC2021.05 | 50.21 | 32.09 | 45.11 | 29.23 | |
| BERT-GED (SRC)backbone=BERT, input=Source only2021.05 | 37.58 | 45.81 | 38.98 | 12.71 | |
| BERT-LMbackbone=BERT2021.05 | 36.97 | 43.42 | 38.1 | 8.59 | |
| BERT-GQEbackbone=BERT2021.05 | 36.05 | 43.53 | 37.33 | 10.18 |