Who Had the Upper Hand? Ranking Participants of Interactions Based on Their Relative Power

International Joint Conference on Natural Language Processing, pages 365–373, Nagoya, Japan, 14-18 October 2013. Who Had the Upper Hand? Ranking Participants of Interactions Based on Their Relative Power Vinodkumar Prabhakaran Columbia University New York, NY, USA vinod@cs.columbia.edu Ajita John Avaya Labs Basking Ridge, NJ, USA ajita@avaya.com Dor´ee D. Seligmann Avaya Labs Basking Ridge, NJ, USA doree@avaya.com Abstract In this paper, we present an automatic sys- tem to rank participants of an interaction in terms of their relative power. We find several linguistic and structural features to be effective in predicting these rank- ings. We conduct our study in the domain of political debates, specifically the 2012 Republican presidential primary debates. Our dataset includes textual transcripts of 20 debates with 4-9 candidates as partic- ipants per debate. We model the power index of each candidate in terms of their relative poll standings in the state and na- tional polls. We find that the candidates’ power indices affect the way they inter- act with others and the way others inter- act with them. We obtained encouraging results in our experiments and we expect these findings to carry across to other gen- res of multi-party conversations. 1 Introduction Recently, there has been a rapid increase in so- cial interactions being stored on the World Wide Web. In addition to those interactions that are in- herently online such as discussion forums and so- cial networks, offline interactions such as broad- cast events, debates and speeches are also cap- tured in real time and stored online in repositories such as YouTube and news media outlets. This growing mass of public data representing various modes of interactions enables researchers to com- putationally analyze social interactions at a scale which was not feasible previously. Within the field of analyzing online social interactions, there is a growing interest to study how the power or sta- tus difference between participants is reflected in the various facets of interactions and if it can be detected using computational means (Diesner and Carley, 2005; Rowe et al., 2007; Bakshy et al., 2011; Bramsen et al., 2011; Biran et al., 2012; Danescu-Niculescu-Mizil et al., 2012). When people interact with one another, there is often a power differential that affects the way they interact. This differential may be drawn from a multitude of factors such as social status, au- thority, experience, age etc. Identifying the domi- nant participants of an interaction through a power ranking system could have various applications. It could help improve effectiveness of advertise- ments within online communities. For example, targeting an advertisement to powerful and influ- ential members within an online community might increase its effectiveness and reach to the commu- nity members. Power analysis can also help in in- formation retrieval systems. Revealing power dy- namics within stored interactions could be useful in determining relevance for a user with informa- tion needs. For example, a user may want to limit his search to posts authored by interactants with higher power. Power analysis may also aid intelli- gence agencies to …

Content truncated to 500 words