Massive Meme Identification and Popularity Analysis in Geopolitics

Abstract

Geopolitics is a long-lasting key issue for governments and nations to assess the international political landscape. The great proliferation of social media in recently years have provided a new avenue to make such political actions in a data driven manner. As the information consumption ability of human is limited, there demands an automatic approach to effectively identify and trace the bursts continuously emerging on social media platforms. Existing studies focusing on named entities recognition or topic detection could provide useful insights for analyzing events that are already known, yet they are incapable of identifying timely emerging trending catchphrase or topics, or memes in general.To tackle with this issue, we elaborate a framework to identify online memes and trace their future dynamics. This framework identify memes based on their independency with regard to the context, and aggregate literal variants of a same meme together into a memeplex with a newly proposed MemeMesh algorithm. Evaluation results on a large scale Twitter dataset suggest that the framework could identify geopolitical memes effectively. Further exploration on meme popularity factors reveals that popularity memes tend to generate more variants during their diffusion, and establish their dominance by attracting a large volume of active users engaging in their diffusion. Causality analysis between meme diversity and user volume suggests that high diversity of meme variants can attract more users involved in spreading a meme at the initial, but these users seldom regenerate more variants in the later time.