The Pursuit of Being Heard: An Unsupervised Approach to Narrative Detection in Online Protest

Protests and mass mobilization are scarce; however, they may lead to dramatic outcomes when they occur. Social media such as Twitter has become a center point for the organization and development of online protests worldwide. It becomes crucial to decipher various narratives shared during an online protest to understand people's perceptions. In this work, we propose an unsupervised clustering-based framework to understand the narratives present in a given online protest. Through a comparative analysis of tweet clusters in 3 protests around government policy bills, we contribute novel insights about narratives shared during an online protest. Across case studies of government policy-induced online protests in India and the United Kingdom, we found familiar mass mo-bilization narratives across protests. We found reports of on-ground activities and call-to-action for people's participation narrative clusters in all three protests under study. We also found protest-centric narratives in different protests, such as skepticism around the topic. The results from our analysis can be used to understand and compare people's perceptions of future mass mobilizations.

I. Introduction

Social media has become integral to various social movements, and protests due to easy information dissemination and wider public reach [1]–[5]. Irrespective of the different socio-economic circumstances or political agendas, the var-ious online protests share similar morphological features in using social media for self-organization and obtaining a more significant number of participants [[6]]. Using a hash-tag to build a collective narrative makes Twitter one of the prime spots for conducting protest [7]. While Twitter enables a broad reach of the protest, a fine-grained analysis of various narratives present within a protest setting may also help decipher the people's perception and shed light on people's will and social protest's overall focus.