Wolof Ajami
First Wolofal HTR/OCR dataset
Official site consultedAI: benchmarks and research
Last verified: 1 October 2026
First dataset of manually segmented and transcribed Wolof manuscripts in Ajami script for handwritten text recognition (HTR) and OCR, presented at LREC 2026. It includes images of 20th-century manuscripts, segmentation, ALTO transcriptions and a keyboard tool.
Services
- Wolof manuscript images
- Manual transcriptions in ALTO format
- Wolof Ajami transcription keyboard
Access requirements
Dataset described as open source; tested models reach 61 to 81% character error rates.
Sources
- Wolof Ajami (LREC 2026) (new tab), verified on 1 October 2026Official site consulted