Argumentation Schemes in German Transcribed Natural Speech
Argumentation mining is an area which has recently gained momentum across disciplines,
in particular in the areas of Artificial Intelligence (AI), computer science and linguistics. The general aim is to identify, annotate and eventually interpret argumentative structures with automatic means – a task that requires a sound theoretical basis in order to arrive at a meaningful computational analysis and representation. One strand of research is concerned with the classification of argumentative structures into argumentation schemes, templates on the types of arguments that allow for inferences in the discourse (Toulmin (1958); Walton et al. (2008), inter alia). To date, only a small amount of research has been done on analyzing actual discourse with respect to argumentative structure (Reed and Walton, 2007; Budzynska et al., 2014) and there is no work on investigating argumentation schemes for transcribed natural speech multilogs, i.e. a multi-party conversations, in German. The collaboration with Chris Reed, University of Dundee, is a first attempt to establish a set of argumentation schemes that is valid for German multilogs, pairing the extensive knowledge of Chris Reed with respect to argumentation theory and mining across disciplines and the (computational) linguistic work on extracting argumentative structures in Konstanz. This effort will pave the way for a theoretically as well as computationally well-motivated mining of arguments in German transcribed natural speech.