Date created: | Last Updated:

Category: Procedure

Description: In the analysis of argumentative discourse, the structure of arguments, which involves the combination of claims and premises, presents intricate challenges of identification and classification. This paper reports on a project aimed at providing data for comparing argument recognition both in machine learning and human discourse processing. The paper has two aims: First, we detail our approach to annotating two argumentative corpora, utilizing a function-based classification scheme complemented by semantic templates for its operationalization. By describing this method in detail, we offer comprehensive guidelines for a classification of arguments, which, in our experience, facilitate annotation. Second, we provide a reflective discus- sion on the procedure and its application to the intricate variety of corpus data. We illustrate our procedure by including corpus attestations for all relevant categories, which sheds light on the practical implications of the approach and on the notorious problems of argument detection as well as proposed solutions. In this way, the documentation of our project has the potential to streamline argument-mining endeavors and, overall, improve the understanding of argumentative discourse.