yiyic/

ukp_m5

This dataset is for fine-tuning language models for zero-shot relation classification for Creoles. See https://github.com/hclent/CreoleVal/tree/main/nlu/relation_classification. The dataset is from UKP relation classification, and it is split with 5 unseen relations in test dataset, while train dataset have 108 relations.