CroSloMet: Croatian-Slovene datasets for metaphor detection and metaphor name generation

CroSloMet is a collection of four Croatian and Slovene datasets for metaphor detection and metaphor name generation tasks. It contains over 1,120 metaphorical and 1,120 literal sentences in Croatian and Slovene, which are grounded in the MetaNet.HR framework. Each sentence is annotated with the corresponding conceptual metaphor, linguistic multi-word expression, canonical forms, and literal usage. Two datasets are used for metaphor detection and two for metaphor name generation. In the metaphor detection sets, the text contains an expression marked with asterisks, and the task is binary: the expression is either metaphorical (1) or literal (0). In the metaphor name generation sets, the text contains an expression marked with asterisks, and the task is to generate either the English name of the metaphor or the Croatian name in the Croatian data (or "/" if the expression is literal). For the Slovenian data, only the English metaphor name is used.