Abstract

Abstract This paper addresses the task of measuring Twitter social attributes that can be used for detecting patterns that show user participation in public debates in South African. We propose a method that leverages observable information on Twitter such as use of language, retweeting user behaviour, and the relationship between topics and the user social network graph. Our experimental results suggest high degrees of citizen participation: people in an otherwise multilingual country tweet in a dominant language; there is more original commentary and interactive discussion; and topics often span natural online communities.