Entropy in network community as an indicator of language structure in emoji usage: A twitter study across various thematic datasets

Abstract

Emojis are emerging as an alternative way to interact and communicate online, and their large-scale adoption has the potential to reveal distinct patterns of human communication and social interactions. In this work, we investigate the hypothesis that emojis are a kind of language. By building networks of emoji co-occurrence, we examine the diversity of the community structure of such networks with regards to predefined categories of emojis. Using four different techniques of community detection, we validate our hypothesis on six Twitter datasets: five from specific topics and one random dataset. Our results demonstrate that the community structure of emojis is more diverse when they are used in non-random topics such as politics and sports, and that Stochastic Block Models appears to extract communities with higher diversity.