In this study we provide a comprehensive empirical evaluation of numerous public datasets commonly employed in automated hate speech classification. Through rigorous empirical analysis, we present compelling evidence that sheds light on the limitations inherent in the current hate speech datasets used in supervised hate speech classification tasks. Furthermore, we offer a range of statistical analyses to elucidate the weaknesses and strengths inherent in these datasets.
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4504059