Using interaction networks to map communities on Twitter

Abstract

This article summarizes our work using a network-based approach to mapping the main constituencies discussing specific topics on Twitter. The approach gives researchers unique insight into the main groups involved and their agendas. While the individual pieces of the methodology are not new, the way in which we combine them will be novel to many market researchers. By connecting Twitter users that interact with each other into an “interaction network” or “conversation map” and using community detection algorithms to isolate distinct groups, we are able to identify the main constituencies discussing a specific topic such as a national election. Once the main constituencies have been identified, it is possible to profile them in more detail such as in terms of their demographics, their influencers, and the type of content that resonates with them. This article specifically focuses on roughly one million tweets about the 2014 South African national election to illustrate our approach. The authors believe that the approach described has wide-ranging applications and that it can be used to give researchers unprecedented insight into the public discourse surrounding specific topics and events. This article is adapted from an earlier article presented at the 2015 Southern African Marketing Research Association’s (SAMRA) annual conference.