n this paper, we present the annotation guidelines we defined for annotating arguments in political debates. In our guidelines, we consider each argument as being composed of a claim and one or more premises. The annotation process has started with defining the guidelines for three annotators containing examples from the data, and continued as cyclic process of evaluation and revision on the annotation to resolve the ambiguities in the guidelines. In this paper, we briefly discuss the resulting annotated dataset and give some examples of the annotation scheme. The quality of the annotated dataset is assessed by computing inter-annotator agreement using Krippendorf’s α coefficient on a portion of the dataset.