Exploring Reasoning Schemes: A Dataset for Syllogism Figure Identification

Argument mining aims at extracting structured arguments from texts. Both argument schemes and their structures are studied. A syllogism is a traditional argument scheme and its figures are important in classical logic. Existing research has not yet investigated syllogism figures. Here, we fill this gap by presenting a study on automatic identification of syllogism figures. We prepared a novel dataset of 8.6k syllogisms. We annotated their figures and carried out identification tasks using both supervised and weakly-supervised approaches. Experimental results show that both approaches are adequate.