M-STANCE is a multilingual, multi-target and multi-cultural stance detection (SD) dataset. It covers social media posts from 2014 and 2019 related to the migration crisis in EU and covers four languages: English, German, Italian and Polish. In this dataset, we first propose a list of broad targets, which are further subcategorized into fine-grained targets. The fine-grained targets cover the fine-grained aspects of each broad targets. For instance, "refugees" is the fine-grained target of broad target "migrants". The targets of the dataset is annotated by LLM and the stance is annotated by humans.

We also conducted cross-cultural annotation between polish and English-German on both directions. In cross-cultural annotation, the posts in the source language are translated to the target language and annotated by speakers of the target language. The resulting cross-culturally annotated dataset can be found under the directory x-culture.