An Object-oriented Model of Role Framing and Attitude Prediction

According to Rashkin et al. (2016) and other researchers, a text not only reveals the attitudes of the discourse referents dealt with in the text, but it also indicates how they are framed by the text author. Do they play a positive or negative role in the exposed narrative? We call this the author’s role framing. While the attitudes tell us who is for or against who (pro and con relations), the role framing indicates who is a beneficiary or a victim etc. of the situation described. Both, attitude and role framing depend on the factuality status of the described situations. For instance, if A hopes that B helps C, than A has a positive attitude towards C, but neither is C is beneficiary nor is there an attitude of B towards C. The reason is that hope is a non-factive verb and help, thus, non-factual. If we were told that A might hope that B helps C, nothing follows at all, since now even hope in non-factual. Not only verbs like hope, but also nouns like destruction and predicatively used adjectives like e.g. angry in Trump is angry that our neighbors are hurting Wisconsin and other border states indicate attitude. Here, Trump and Canada (neighbors) are adversaries, but Trump is pro Wisconsin, which is framed as a victim. Moreover, constructions like It is sad that .. clearly reveal the stance of the author towards an event and as a consequence towards the participants of this event. So in the end, we are able to derive the author’s stance, the proclaimed attitudes among referents and the role framing that comes with a text. We introduce an object-oriented model realized in Python to deal with these topics. An empirical evaluation is presented concerning predicatively used adjectives.