Multi-view Personality Profiling Based on Longitudinal Data

Personality profiling is an essential application for the marketing, advertisement and sales industries. Indeed, the knowledge about one's personality may help in understanding the reasons behind one's behavior and his/her motivation in undertaking new life challenges. In this study, we take the first step towards solving the problem of automatic personality profiling. Specifically, we propose the idea of fusing multi-source multi-modal temporal data in our computational "PersonalLSTM" framework for automatic user personality inference. Experimental results show that incorporation of multi-source temporal data allows for more accurate personality profiling, as compared to non-temporal baselines and different data source combinations.