olaverse/

diacnet-1.1-train

Ready-to-train (input, target) pairs for diacnet-1.1, a 10-language diacritic-restoration model. This is the final stage of the DiacNet 1.1 data family — sentences from diacnet-1.1-corpus with all training augmentations applied, including gloss conditioning from diacnet-1.1-gloss.

	The DiacNet 1.1 data family




	

Dataset Role

diacnet-1.1-corpus Clean diacritized sentences (FineWeb-2 + Wikipedia, density-filtered)

diacnet-1.1-gloss… See the full description on the dataset page: https://huggingface.co/datasets/olaverse/diacnet-1.1-train.