Effectiveness of Counter-Speech against Abusive Content: A Multidimensional Annotation and Classification Study