ImageMet is a synthetically generated and human-validated dataset that was constructed using a systematic methodology grounded in the source-target domain mappings established by Lakoff and Johnson (1980) and on the linguistic metaphor annotation guidelines developed by Pragglejaz (2007); Steen et al. (2010); Sanchez-Bayona and Agerri (2022).
ImageMet dataset has a size of 639 instances. Each instance consists of:
the visual metaphor (image_path),
the source and the target of the conceptual… See the full description on the dataset page:
https://huggingface.co/datasets/AndreaJaunarena-Cayetano/ImageMet.
