This dataset consists of digitized Balochi printed books licensed by publishers and authors, including rare lithographed prints in Arabic-derived script. It contains original page scans, OCR-generated text, and native-expert, cross-verified ground-truth transcriptions.
Potential Use Cases
- Balochi OCR/VLM Model Training (Image + Text Pairs): Page scans paired with native-verified ground-truth transcriptions can be used to train or fine-tune OCR and vision-language models (VLMs) to recognize the complex character forms found in rare lithographed prints.
- Balochi Text Resource Building (Text-Only, No Images): The verified ground-truth text can be extracted independently of the page images and used as a language resource for low-resource NLP work, such as corpus building and text normalization for Balochi.