Token Level Quality Estimation on CoNLL ann. 1 2014 (test)
76.03PrecisionBERT-VERNet
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BERT-VERNetScenario=Hypothesis, Input Encoding=Source, hypothesis pairs2021.05 | 76.03 | 34.02 | 60.97 | |
| ELECTRA-VERNetScenario=Hypothesis, Input Encoding=Source, hypothesis pairs2021.05 | 74.8 | 39.26 | 63.33 | |
| BERT-GED (HYP)Scenario=Hypothesis, Input Encoding=Source, hypothesis pairs2021.05 | 74.28 | 34.2 | 60.17 | |
| BERT-GED (JOINT)Scenario=Hypothesis, Input Encoding=Source, hypothesis pairs2021.05 | 71.15 | 38.3 | 60.73 | |
| BERT-VERNetScenario=Source, Input Encoding=Source, hypothesis pairs2021.05 | 62.64 | 30.62 | 51.8 | |
| ELECTRA-VERNetScenario=Source, Input Encoding=Source, hypothesis pairs2021.05 | 62.5 | 35.61 | 54.3 | |
| BERT-GED (JOINT)Scenario=Source, Input Encoding=Source, hypothesis pairs2021.05 | 60.79 | 27.33 | 48.83 | |
| BERT-GED (SRC)Scenario=Source, Input Encoding=Source sentences only2021.05 | 59.84 | 27.11 | 48.2 |