Framing of Political Figures Through Titles and Through Background Information
In this talk I first present a quantitative investigation of entity framing in English and German. We labeled tweets for stance towards politicians and find that the formality of naming forms for targets correlates positively with stance. This suggests a primarily status-indicating function of names and titles in such data. On the German corpus we find that this status-indicating function is much weaker in tweets from left-leaning users that from right-leaning users.
In a second part, I introduce the task of detecting Informational Bias, which is the framing of entities through selective inclusion of background information, and two approaches to this task that we are currently exploring.