AraCrypto-ML is a 10,000-post multi-label corpus covering Arabic Facebook cryptocurrency discussions from 2014 to 2024. The dataset uses Fanar-2-27B to annotate six categories including market commentary and religious framing, achieving a macro kappa of 0.74 against gold labels.
HOW THIS AFFECTS YOU
●
researcherYou can use this dataset to train models on specialized, multi-label Arabic social media text.