Social Media Attributions in the Context of Water Crisis

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 1402–1412, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 1402 Social Media Attributions in the Context of Water Crisis Rupak Sarkar♣∗ Sayantan Mahinder♥∗ Hirak Sarkar♦ Ashiqur R. KhudaBukhsh♠ ♣Maulana Abul Kalam Azad University of Technology ♥Independent Researcher ♦University of Maryland ♠Carnegie Mellon University rupaksarkar.cs@gmail.com, sayantan.mahinder@gmail.com, hsarkar@cs.umd.edu, akhudabu@cs.cmu.edu Abstract Attribution of natural disasters/collective mis- fortune is a widely-studied political science problem. However, such studies typically rely on surveys, expert opinions, or external signals such as voting outcomes. In this paper, we ex- plore the viability of using unstructured, noisy social media data to complement traditional surveys through automatically extracting attri- bution factors. We present a novel prediction task of attribution tie detection of identifying the factors (e.g., poor city planning, exploding population etc.) held responsible for the cri- sis in a social media document. We focus on the 2019 Chennai water crisis that rapidly es- calated into a discussion topic with global im- portance following alarming water-crisis statis- tics. On a challenging data set constructed from YouTube comments (72,098 comments posted by 43,859 users on 623 videos relevant to the crisis), we present a neural baseline to identify attribution ties that achieves a reason- able performance (accuracy: 87.34% on attri- bution detection and 81.37% on attribution res- olution). We release the first annotated data set of 2,500 comments in this important domain1. 1 Introduction Water crisis is one of the pressing current environ- mental challenges. More than a billion people do not have access to clean drinking water, and every year nearly two million children die from water borne diseases (Watkins, 2006). One-third of the world’s most extensive groundwater systems are ∗ Rupak Sarkar and Sayantan Mahinder are equal- contribution first authors. Ashiqur R. KhudaBukhsh is the corresponding author. 1Code and data are publicly available at https://www. cs.cmu.edu/˜akhudabu/WaterCrisis.html. under severe stress (Richey et al., 2015). The fore- casts look even more grim; nearly two-thirds of the world population could be water stressed by 2025 (Seckler et al., 1999). While the crisis has reached an alarming level far and wide, India is listed as one of the major at-risk countries (Rost et al., 2008). In June 2019, the longstanding Chen- nai water crisis (WashingtonPost, 2019) escalated into an international talking point, revealing alarm- ing statistics of the water crisis in India looming in near future. In this context, we define a new task of inferring attribution ties through large scale analysis of relevant social media discussions. Our main contributions in this paper are the following. Social: Apportioning attribution for a collective crisis or misfortune still remains a challenge in so- cial science, despite the presence of a large body of political science literature on retrospective voting (see, e.g., Ferejohn 1986; Peffley 1984) or psycho- logical literature on attribution (see, e.g., Shaver 2012). Prior social science literature (Griffin et al., 2008) primarily relies on traditional surveys for attribution analysis. Unlike traditional surveys, so- cial media analyses are vastly cheaper, have …

Content truncated to 500 words