Predicting temperament from twitter data

Abstract

Temperament is an innate psychological characteristic associated with how we relate with the world. This feature is often used to direct careers, manage conflicts, develop leadership, improve teaching, etc. The data generated by social media users represent user behavior facing the various situations of everyday life. With this, machine learning techniques can be used to infer the tem-perament, as is already done in the vast research on sentiment analysis and the growing research on personality prediction. This paper proposes a framework for temperament classification according to the theory of psychologist David Keirsey. Our results present an accuracy higher than 70% for the Artisan and Guardian types.