Communication scientists suggest that public opinion on societal and 
political issues is formed not only through explicitly opinionated 
discussion, but also implicitly through framing, defined as the 
selective mentioning of certain aspects of topics to promote a 
particular interpretation. There exists a small but growing number of 
efforts to automatically detect framing of events, to which we want to 
contribute work focusing on the framing of entities. Our first aim is to 
demonstrate that the selection or omission of objectively verifiable 
attributes such as name, age and occupation can contribute to the stance 
of a text. To this end we performed a correlation study on 
sentiment-tagged tweets mentioning the presidents of five G20 countries 
using varying reference forms. We then investigated whether the stance 
of newspaper articles on Donald Trump during his 2016 presidential 
campaign can be predicted by the presence of mentions of his previous 
occupations. In this talk I will present the results of these two 
experiments, as well as discuss the initial stages of future work.