Machine Translation on Adapted-Domain (AD) (val)
23.35BLEUSeq2Seq_ID + Seq2Seq_AD
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Seq2Seq_ID + Seq2Seq_ADSchedule=Fine-tuning (Parallel), Condition=Hypothetical supervised adaptation2022.03 | 23.35 | 0.836 | |
| BackTr_ADSchedule=Fine-tuning2022.03 | 22.98 | 0.834 | |
| Seq2Seq_ID + BackTr_ADSchedule=Fine-tuning (Parallel)2022.03 | 21.47 | 0.831 | |
| Seq2Seq_ID + Seq2Seq_ADSchedule=Pre-training (Parallel), Condition=Hypothetical supervised adaptation2022.03 | 18.55 | 0.813 | |
| Seq2Seq_IDSchedule=Fine-tuning2022.03 | 17.62 | 0.808 | |
| Seq2Seq_ID + Seq2Seq_ADSchedule=Pre-training (Sequential), Condition=Hypothetical supervised adaptation2022.03 | 17.37 | 0.816 | |
| Seq2Seq_ID + BackTr_ADSchedule=Pre-training (Parallel)2022.03 | 16.99 | 0.817 | |
| Seq2Seq_ID + BackTr_ADSchedule=Pre-training (Sequential)2022.03 | 15.01 | 0.805 | |
| Seq2Seq_ID + Denois_ADSchedule=Fine-tuning (Parallel)2022.03 | 13.37 | 0.782 | |
| Seq2Seq_IDSchedule=Pre-training2022.03 | 5.34 | 0.738 |