chrismazii/kinycomet_dataset
Domain:
natural language processingRecord type:
dataset
Creator:
chr
Host:
This dataset contains 4,323 human-annotated translation quality assessments for Kinyarwanda ↔ English translation pairs. It was specifically created to train the KinyCOMET model, addressing the critical need for reliable automatic evaluation of Kinyarwanda translations.
Total Samples: 4,323 translation pairs with quality scores