Open Conference Systems, CLADAG2023

Font Size: 
Cluster analysis for the study of online visual communication
Matteo Magnani, Matias Piqueras, Alexandra Segerberg, Davide Vega, Victoria Yantseva

Last modified: 2023-07-10


Several online data analysis tasks such as the detection of inauthentic coordinated behaviour and the study of online polarisation and political persuasion require the summarisation of large numbers of social media posts. Therefore, cluster analysis is a fundamental tool for the study of online communication. However, most of the existing computational studies of online communication, including those based on clustering methods, have focused only on thetext contained in social media posts. This, despite the fundamental role played by visual elements, such as images and videos, in the main social media platforms. For example, images often carry an important part of the information contained in a post, are in general associated to increased spreading, and can be shared across languages. In this talk I will first provide an overview of the available approaches to cluster images, also discussing the specificities of clustering social images (that is, images shared on social media) and how to use image clustering as part of a social data science process. Then I will present an approach to cluster online communication networks based on image clustering. I will conclude by showing an application of image-based communication network clustering to study online visual climate change communication.

Full Text: PDF