Measuring political influence during elections using a deep learning approach

Abstract

This contribution introduces a methodology for measuring political influence on Twitter during the 2020 U.S. presidential election campaign. The approach employs deep knowledge scores, which are generated through sentiment analysis of Tweets from users responding to influential users, coupled with an assessment of the strength of their interactions. The deep knowledge scores enable the categorization of three types of Twitter’s users engaging with influential users: influenced users, distrustful users, and connected users. Our approach, structured around a five-layer framework, effectively constructs networks of trust and distrust, and establishes the relationship between fluctuations in trust or distrust levels and the topics discussed by influential users.