Occupational Representativeness in Twitter

Abstract

This paper describes an approach to detect one particular demographic characteristic, occupation (or profession) in Twitter user profiles. In this paper, we show how effective the approach is for estimating occupational population statistics in Australian Twitter by correlating them with real-world population obtained from 2011 Australian census data. We also demonstrate that we can gain more reliable social media insights in the context of occupational representativeness in Twitter if a non-standard occupation name is mapped into a standard occupation name. To our knowledge, this is the first attempt to build a machine learning model that automatically identifies linguistically noisy or open-ended occupations in Twitter, resulting in more reliable occupational population.