Modeling cross-platform narrative templates: a temporal knowledge graph approach
Ridwan Amure, Nitin Agarwal
Abstract
Abstract Over the past decade, online social media has grown in size, features, and complexity, providing users with increased satisfaction and prompting many to maintain accounts across multiple platforms. Information actors have also taken advantage of this environment, using cross-platform dynamics to amplify content’s reach and target specific audiences strategically. As these actors will likely continue exploiting social media, we argue that it is crucial to model cross-platform narratives effectively and identify the patterns-or templates defined in this research-they use to propagate different narratives. To address these challenges, we leverage temporal knowledge graphs to model the relationships between cross-platform narratives, extract temporal communities representing macro-narratives, and apply sequential mining to uncover various narrative templates. These templates reveal the patterns various actors use to spread different narratives across various social media platforms. An analysis of 4,817 Instagram posts, 2,560 TikTok posts, 11,134 X posts, and 7,327 YouTube posts, demonstrates the efficacy of this approach in identifying the templates preferred by Pro-Taiwan and Pro-China actors in the Asia–Pacific political landscape. We identified two groups of narrative templates based on confidence and support. Our further analysis uncovers which templates were favored by Pro-Taiwan and Pro-China supporters.