Code-switched text synthesis on Hi-En
28.88BLEUFine-tuned PMMTM on all language pairs (augment-MMT)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Fine-tuned PMMTM on all language pairs (augment-MMT)Type=Supervised2023.05 | 28.88 | 55.1 | |
| Fine-tuned PMMTM on all language pairs (mBART50-MMT)Type=Supervised2023.05 | 27.93 | 54.81 | |
| Gupta et al. (2020)Type=Supervised2023.05 | 21.55 | 28.37 | |
| GLOSS (augment-MMT + prefix)Type=Zero-shot transfer2023.05 | 12.16 | 36.94 | |
| Machine TranslationType=Unsupervised2023.05 | 9.87 | 24.26 | |
| GLOSS (augment-MMT + adapter)Type=Zero-shot transfer2023.05 | 8.61 | 30.39 | |
| GLOSS (mBART50-MMT + prefix)Type=Zero-shot transfer2023.05 | 7.51 | 29.82 | |
| Translate, Align, then SwapType=Unsupervised2023.05 | 6.61 | 24.9 | |
| Copy InputType=Unsupervised2023.05 | 5.22 | 24.2 | |
| GLOSS (mBART50-MMT + adapter)Type=Zero-shot transfer2023.05 | 4.09 | 22.02 | |
| Fine-tuned PMMTM on available language pairsType=Zero-shot transfer2023.05 | 3.93 | 22.22 |