nrl-ai/

vn-diacritic-train

Two register-distinct training corpora used to fine-tune Vietnamese diacritic-restoration models in the nom-vn project. Each row is a JSONL record: {"input": "diacritic-stripped text", "target": "correctly diacriticized original"}

Inputs are produced by nom.text.strip_diacritics on the target. Both fields are NFC-normalized.

	Configs




	
	wiki_500k

500K (input, target) pairs from… See the full description on the dataset page: https://huggingface.co/datasets/nrl-ai/vn-diacritic-train.