TwiSP: a framework for exploring polarized issues in Twitter

GABRIELA DIAZ; CARLOS CHESÑEVAR; ELSA ESTÉVEZ; ANA MAGUITMAN

Belo Horizonte

Congreso; 16th International Conference on Theory and Practice of Electronic Governance; 2023

ICEGOV Steering Committee

Social and political polarization has become a dramatically intensifying force that is having a huge impact on political discourse, public policies and electoral outcomes in the 21st century. Twitter is a social media platform that mirrors to a large extent the sociological notion of public opinion, and has notably fueled these polarization dynamics worldwide. A proper understanding of how different issues become polarized in Twitter and their interrelationship is therefore crucial for the development of effective policies and governance strategies in our democracies. This paper introduces TwiSP, a framework for analyzing polarization on controversial topics in Twitter. TwiSP utilizes a combination of two cutting-edge machine learning techniques: stance detection for identifying attitudes and perspectives and BERTopic for topic modeling. The outcome of TwiSP is a visual tree-like representation of all tweets related to conflicting topics (rooted in a particular topic), contrasting their relationship using different colors to denote the degree of polarization. As a case study, we show how the TwiSP framework can be used for analyzing polarized issues in the context of the COVID-19 vaccine, exploring the resulting degree of polarization and the key topics driving it. The results reveal the diversity of opinions and the presence of highly polarized clusters in social media discussions. We contend that the TwiSP framework provides a novel and valuable tool for decision makers, helping them to recognize contentious issues behind the dynamics of polarization and ultimately identifying potential opportunities for bridging divides.