Detecting Influential Twitter Accounts based on Sensitive Topics with Graph Modelling

Sensitive content on Twitter, often accompanied by abusive comments, significantly impacts public perception, potentially inciting violence and causing psychological distress. Previous research has typically identified influential accounts using single metrics, such as follower count or retweet frequency, which do not capture the complexity of influence within social networks. The impact of accounts in spreading sensitive content also varies depending on the distribution patterns and shared content volume, highlighting the need for a content-based approach to understanding these differences. This study addresses these limitations by introducing a multidimensional approach that combines content analysis with social network metrics such as degree centrality, betweenness centrality, closeness centrality, and eigenvector centrality. Different from traditional ranking methods based solely on centrality metrics, this approach demonstrates that the most influential accounts in a network are only sometimes the most active or those with the highest follower counts. Our findings reveal clear ranking differences between structurally influential accounts and those actively engaged in disseminating sensitive content, emphasizing the need for a multidimensional perspective in influence analysis.