Affilitiation: Professor of Computational Linguistics at University of Antwerp


Detecting author personality from text

In computational stylometry, psychological and sociological properties of people are assigned on the basis of texts written by these people. One of the properties under investigation at CLiPS is personality. We describe recent empirical work on assigning MBTI and “Big Five” type personality profiles to authors of text using text analysis and machine learning methods. We show that it is a hard problem to unravel the interaction of indicators of different author properties (notably age and gender) with personality indicators, and show that a new type of corpus is needed for solving this problem. We also report on progress developing such a corpus, and go into applications of personality assignment, most notably in human resources and in marketing.