VMC-P (VU Amsterdam Metaphor Corpus with Paraphrases) is a dataset, developed for end-to-end metaphor interpretation. It consists of 10,716 textual sequences (6,653 sequences for training, 2,063 sequences for validation, and 2,000 sequences for testing) with manual annotations on the token level. These annotations specify the metaphricity (such as metaphorical or literal) and provide paraphrases of metaphorical lexical units… See the full description on the dataset page: https://huggingface.co/datasets/RuiMao1988/VMC-P.
