Advancing Stance Detection of Political Fan Pages: A Multimodal Approach

The evolution of political campaigns is evident with the ascent of social media. Ideological beliefs are increasingly disseminated through political-affiliated fan pages. The interaction between politicians and the general public on these platforms plays a pivotal role in election outcomes. In this study, we utilize a multimodal approach to explore and quantify similarities of ideologies among political fan pages. we employed visualization techniques to demonstrate the political stance of each fan page. To validate our proposal, we concentrated on an analysis of the 2021 national referendums in Taiwan, encompassing a collection of fan pages and their corresponding posts that were related to these referendums. Through a qualitative analysis of the content of these fan pages, the efficacy of our multimodal framework in clustering fan pages according to their respective political ideologies was evaluated. The findings of this study underscore the significant enhancement in the accuracy of stance detection when integrating multiple modalities of data, namely textual content, visual imagery, and user interactions.