codewithnawaB/

pashto-corpus-v1

A large-scale Pashto language corpus for NLP research, covering both Afghan (Kandahari) and Pakistani (Peshawari) dialects.

	Dataset Statistics




	

Split Count

Pre-training documents 2,021,382

SFT instruction pairs 9,250,673

Total 11,272,055

	Dialect Distribution




	

Dialect Count %

Peshawari (Pakistani) 1,992,924 98%

Afghan (Kandahari) 28,458 1%

Mixed / Unknown 0 0%

BBC… See the full description on the dataset page: https://huggingface.co/datasets/codewithnawaB/pashto-corpus-v1.