Evaluating Content-Independent Features for Personality Recognition

Abstract

This paper describes our submission for the WCPR14 shared task on computational personality recognition. We have investigated whether the features proposed by Soler and Wanner (2014) for gender prediction might also be useful in personality recognition. We have compared these features with simple approaches using token unigrams, character trigrams and liwc features. Although the newly investigated features seem to work quite well on certain personality traits, they do not outperform the simple approaches.