Align Voting Behavior with Public Statements for Legislator Representation Learning

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 1236–1246 August 1–6, 2021. ©2021 Association for Computational Linguistics 1236 Align Voting Behavior with Public Statements for Legislator Representation Learning Xinyi Mou1, Zhongyu Wei1,2∗, Lei Chen1, Shangyi Ning1, Yancheng He3, Changjian Jiang4∗, Xuanjing Huang5 1School of Data Science, Fudan University, China 2Research Institute of Intelligent and Complex Systems, Fudan University, China 3Platform and Content Group, Tencent, China 4School of International Relations & Public Affairs, Fudan University, China 5School of Computer Science, Fudan University, China {xymou20,zywei,chenl18,syning18}@fudan.edu.cn collinhe@tencent.com, {Changjian,xjhuang}@fudan.edu.cn Abstract Ideology of legislators is typically estimated by ideal point models from historical records of votes. It represents legislators and legis- lation as points in a latent space and shows promising results for modeling voting behav- ior. However, it fails to capture more specific attitudes of legislators toward emerging issues and is unable to model newly-elected legisla- tors without voting histories. In order to miti- gate these two problems, we explore to incor- porate both voting behavior and public state- ments on Twitter to jointly model legislators. In addition, we propose a novel task, namely hashtag usage prediction to model the ideol- ogy of legislators on Twitter. In practice, we construct a heterogeneous graph for the leg- islative context and use relational graph neural networks to learn the representation of legisla- tors with the guidance of historical records of their voting and hashtag usage. Experiment re- sults indicate that our model yields significant improvements for the task of roll call vote pre- diction. Further analysis further demonstrates that legislator representation we learned cap- tures nuances in statements. 1 Introduction Modeling the behavior of legislators is one of the most important topics of quantitative political sci- ence. Existing researches largely rely on roll call data, i.e. historical voting records, to estimate the political preference of legislators. The most widely used approach for roll call data analysis is ideal point model (Clinton et al., 2004) that rep- resents legislators and legislation as points in a one-dimension latent space. Researchers enhance ideal point model by incorporating textual infor- mation of legislation (Gerrish and Blei, 2011; Gu ∗Corresponding author. Figure 1: An illustration of correspondence of vote be- havior and public statements on Twitter. Supporters of the abortion-banning legislation frequently mention the tag life while opponents focus on choice. et al., 2014; Kraft et al., 2016) and report positive results for roll call vote prediction. Although roll call data is the major resource for legislator behavior modeling, it has two limitations. Firstly, it fails to uncover detailed opinions of legis- lators towards legislative issues. Therefore, we have no clue about the motivation behind their voting. Secondly, it is unable to model the be- havior of newly-elected legislators because their historical voting records are not available (i.e., cold- start problem). Meanwhile, researchers explore to use public statements to characterize the ideology of legislators with the guidance of framing the- ory (Entman, 1993; Chong and Druckman, …

Content truncated to 500 words