Institute of Formal and Applied Linguistics

"Framing" in communication refers to the mechanisms of selecting aspects of a perceived reality and making them more salient, in order to promote a particular problem definition, causal interpretation, moral evaluation, and/or treatment recommendation (Entman 1993). As such it is (trivially) a necessity in many communication events, and at the same time an instrument to be intentionally used in political discourse. In this talk, we work with a new corpus of 500 "Nature" and "Science" editorials that mention climate change and have been labeled by climate scienctists for 8 different frame categories. We first present a series of supervised classification experiments on this highly imbalanced dataset. In the second part, the literature on approaches to framing in NLP and in the social sciences is briefly reviewed, and directions for alternative approaches to the Nature/Science corpus are explored.