Understanding the Language of Political Agreement and Disagreement in Legislative Texts

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 5358–5368 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 5358 Understanding the Language of Political Agreement and Disagreement in Legislative Texts Maryam Davoodi Purdue University mdavoodi@purdue.edu Eric Waltenburg Purdue University ewaltenb@purdue.edu Dan Goldwasser Purdue University dgoldwas@purdue.edu Abstract While national politics often receive the spot- light, the overwhelming majority of legisla- tion proposed, discussed, and enacted is done at the state level. Despite this fact, there is little awareness of the dynamics that lead to adopting these policies. In this paper, we take the first step towards a better understanding of these processes and the underlying dynam- ics that shape them, using data-driven methods. We build a new large-scale dataset, from multi- ple data sources, connecting state bills and leg- islator information, geographical information about their districts, and donations and donors’ information. We suggest a novel task, predict- ing the legislative body’s vote breakdown for a given bill, according to different criteria of interest, such as gender, rural-urban and ide- ological splits. Finally, we suggest a shared relational embedding model, representing the interactions between the text of the bill and the legislative context in which it is presented. Our experiments show that providing this con- text helps improve the prediction over strong text-based models. 1 Introduction Despite the fact that state-level legislation is rarely discussed, it has a dramatic influence on the every- day life of residents of the respective states. The policies enacted at the state-level touch on all as- pects, from mundane topics, such as trash removal and state mascots, to highly ideologically-charged topics such as education, religious liberties, and health-care access. Moreover, state-legislatures dis- cuss and vote-on significantly more bills than their Federal counterparts, adding up to over 120,000 bills per year (King, 2019). Also, the lack of gen- eral interest, as well as the complexity of the pro- cesses that differ across states, often leads to public disengagement from local politics. This results in decisions being made with little understanding of Republican Democrat b) Competitive c) Inverse-Competitive Yea 20% Nay 80% Yea 70% Nay 30% Yea 55% Nay 45% Yea 25% Nay 75% Yea 30% Nay 70% a) Failed Figure 1: Example of failure and party cleavages. the processes that shape them and how they are likely to influence different demographics. Similarly, most effort directed at understanding political processes using data was directed at the Federal level. In the NLP community, several works looked at analyzing political texts (Iyyer et al., 2014) and the resulting behaviors of legis- lators (Gerrish and Blei, 2011, 2012). The only exception is recent work (Eidelman et al., 2018), predicting whether a bill would pass the prelim- inary stage, legislative committee, to a full-body vote. State-level demographic cleavages: Our goal in this paper is to take a first step towards under- standing the processes and interests that underlie how decisions are passed using data-driven meth- ods. Our main intuition is that the impact of bills on different …

Content truncated to 500 words