Identifying Political Sentiment between Nation States with Social Media

Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 65–75, Lisbon, Portugal, 17-21 September 2015. c⃝2015 Association for Computational Linguistics. Identifying Political Sentiment between Nation States with Social Media Nathanael Chambers, Victor Bowen, Ethan Genco, Xisen Tian, Eric Young, Ganesh Harihara, Eugene Yang Department of Computer Science United States Naval Academy nchamber@usna.edu Abstract This paper describes an approach to large- scale modeling of sentiment analysis for the social sciences. The goal is to model relations between nation states through so- cial media. Many cross-disciplinary appli- cations of NLP involve making predictions (such as predicting political elections), but this paper instead focuses on a model that is applicable to broader analysis. Do cit- izens express opinions in line with their home country’s formal relations? When opinions diverge over time, what is the cause and can social media serve to de- tect these changes? We describe several learning algorithms to study how the pop- ulace of a country discusses foreign na- tions on Twitter, ranging from state-of-the- art contextual sentiment analysis to some required practical learners that filter irrel- evant tweets. We evaluate on standard sentiment evaluations, but we also show strong correlations with two public opin- ion polls and current international alliance relationships. We conclude with some po- litical science use cases. 1 Introduction The volume of text available on social media pro- vides a new opportunity for public policy and po- litical science. Specifically in the area of interna- tional relations, advances in natural language un- derstanding and sentiment analysis may offer new insights into the sentiment of one nation toward another. This paper processes 17 months of Twit- ter data to identify discussions about sovereign states, and it aggregates opinions toward these states from foreign nations. We present a novel application of contextual sentiment with this task, and identify several semi-supervised learning al- gorithms that are needed to address the reference resolution challenge inherent to country names. We present intrinsic evaluations of our learners on labeled datasets as well as four extrinsic politi- cal science evaluations that show strong alignment with our large-scale sentiment extraction. An open question for international policy mak- ers is the extent to which public opinion drives de- cision making. How do military conflicts affect a neutral nation’s relationship? Does public opin- ion shift toward a country after a formal alliance is created, or must popular opinion shift first? These questions are difficult to address due to the lack of measurable data. While polling data can be col- lected, collection beyond a handful of countries is cost prohibitive. This paper hypothesizes that sentiment analysis can be used as a proxy to track international relations between nation states. We describe the largest attempt (over 2 billion tweets) to measure nation state sentiment across hundreds of country pairs. The core challenge to measuring public opin- ion between countries is an accurate algorithm to judge the sentiment of a text toward another na- tion. Unlike traditional sentiment analysis, the general sentiment of the text …

Content truncated to 500 words