The US2016-SCHEMES corpus comprises transcripts of television debates leading up to the 2016 US presidential elections, combined with associated reactions on the Reddit social media platform. The full corpus (97,999 words) is annotated with the argumentative relations of inference, conflict and rephrase, and their illocutionary anchoring in the discourse. A sub-corpus containing the first general election debate between Clinton and Trump (17,190 words) is annotated with argument schemes on the basis of a factorial approach to argument classification called the ‘Periodic Table of Arguments’. This classification is based on three discriminating properties of arguments: first-/second-order; predicate/subject; fact/value/policy propositions. The US2016-SCHEMES corpus serves as an open resource of empirical data on argument schemes – e.g. for the development of argument mining techniques (the automated reconstruction of argumentation).
More details here