Back-transliteration on IndoNLP (Set 2)
0.0903WERFine-tuned BERT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Fine-tuned BERTSystem=Perera et al. (2025)2025.07 | 0.0903 | 0.0215 | 0.9099 | |
| Sinhala BERTSystem=Perera et al. (2025)2025.07 | 0.0917 | 0.0216 | 0.9084 | |
| DL-basedSystem=De Mel et al. (2024)2025.07 | 0.2413 | 0.0789 | 0.4384 | |
| N-gram + Rule-basedSystem=Sumanathilaka et al. (2023)2025.07 | 0.2509 | 0.0678 | 0.7502 | |
| GRU + Rule-basedSystem=Dharmasiri and Sumanathilaka (2024)2025.07 | 0.4808 | 0.1567 | 0.5193 | |
| Rule-basedSystem=De Mel et al. (2024)2025.07 | 0.6809 | 0.2202 | 0.0163 |