Measuring Ideological Proportions in Political Speeches

Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 91–101, Seattle, Washington, USA, 18-21 October 2013. c⃝2013 Association for Computational Linguistics Measuring Ideological Proportions in Political Speeches Yanchuan Sim∗ Brice D. L. Acree† ∗Language Technologies Institute Carnegie Mellon University Pittsburgh, PA 15213, USA {ysim,nasmith}@cs.cmu.edu Justin H. Gross† Noah A. Smith∗ †Department of Political Science University of North Carolina at Chapel Hill Chapel Hill, NC 27599, USA {brice.acree,jhgross}@unc.edu Abstract We seek to measure political candidates’ ideo- logical positioning from their speeches. To ac- complish this, we infer ideological cues from a corpus of political writings annotated with known ideologies. We then represent the speeches of U.S. Presidential candidates as se- quences of cues and lags (filler distinguished only by its length in words). We apply a domain-informed Bayesian HMM to infer the proportions of ideologies each candidate uses in each campaign. The results are validated against a set of preregistered, domain expert- authored hypotheses. 1 Introduction The artful use of language is central to politics, and the language of politicians has attracted consider- able interest among scholars of political commu- nication and rhetoric (Charteris-Black, 2005; Hart, 2009; Deirmeier et al., 2012; Hart et al., 2013) and computational linguistics (Thomas et al., 2006; Fader et al., 2007; Gerrish and Blei, 2011, in- ter alia). In American politics, candidates for of- fice give speeches and write books and manifestos expounding their ideas. Every political season, however, there are accusations of candidates “flip- flopping” on issues, with opinion shows, late-night comedies, and talk radio hosts replaying clips of candidates contradicting earlier statements. Pres- idential candidate Mitt Romney’s own aide infa- mously proclaimed in 2012: “I think you hit a reset button for the fall campaign [i.e., the general elec- tion]. Everything changes. It’s almost like an Etch- a-Sketch. You can kind of shake it up and we start all over again.” A more general observation, often stated but not yet, to our knowledge, tested empirically, is that successful primary candidates “move to the cen- ter” before a general election. The expectation fol- lows directly from long-standing and widely influen- tial theories of political competition that are collec- tively referred to in their simplest form as the “me- dian voter theorem” (Hotelling, 1929; Black, 1948; Downs, 1957). Thus it is to be expected that when a set of voters that are more ideologically concen- trated are replaced by a set who are more widely dispersed across the ideological spectrum, as occurs in the transition between the United States primary and general elections, that candidates will present themselves as more moderate in an effort to capture enough votes to win. Do political candidates in fact stray ideologically at opportune moments? More specifically, can we measure candidates’ ideological positions from their prose at different times? Following much work on classifying the political ideology expressed by a piece of text (Laver et al., 2003; Monroe and Maeda, 2004; Hillard et al., 2008), we start from the as- sumption that a candidate’s choice of words …

Content truncated to 500 words