Understanding subject-based Emoji usage using network science

Abstract

The use of “Emoticons” and “Emojis” in social media as well as most online writing has become the de-facto standard on how to express emotions, feelings, etc. Although there are more that 1,000 emojis, not much has been done to understand the way in which people use these characters. The large set of emojis available brings two questions: (i) How can users make full use of the emojis available? and (ii) Would it be possible to build a recommendation system for emoji usage in text? This paper moves towards a greater understanding of emoji usage by mapping possible relations between these special characters in common text. We look at possible regularities in emoji usages in written, subject-specific, text corpora. We build co-occurrence networks of emoji based on two datasets and show that the structure of these networks are not random and more like a truncated power-law, but more interesting, we show that the structure has similar characteristics despite the text being subject-specific.