Pre-Editorial Normalization on Gold data manual curated (test)
6.7CER (All)ByT5-based normalization model
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| ByT5-based normalization modeltraining_data=Bilingual (Latin and Old French)2026.02 | 6.7 | 6.3 | 7.6 | 20.7 | 19.6 | 23.8 | |
| ByT5 (Schonhardt)training_data=Schonhardt (2025)2026.02 | 64.8 | 75.6 | 33.3 | 92.3 | 104.8 | 55.9 | |
| Llama 3 8B Instructmode=zero-shot2026.02 | 139.9 | 160.4 | 79.9 | 163.7 | 183.1 | 107.3 |