Computational Estimation of Personal Properties From Language

Abstract

Research in natural language and other modalities is starting to shed light on individuals personal properties. Estimating the personal properties of an individual or a group of individuals is the task of detecting different behavioural signals and studying how they correlate with personal properties such as mental health, personality, and emotions. Multiple disciplines such as philosophy, psychology, sociology, and cognitive science focus on defining and detecting personal properties. With the ease of data collection and analysis, studying and analyzing personal properties has become an easier task. Computer science can contribute to this on-going research by building computational models that mimic or predict an individual/group’s personal properties. This kind of research is done through studying two different behavioural signals. In my research I focus on verbal signals by studying how language usage correlates with personal properties. The other behavioural signal is non-verbal, such as body language, number of friends, eating habits, etc. The contribution of my thesis can be broken down to two parts: building tools to estimate a set of individual/group’s personal properties from mainly online posts through their language usage, and comparing the effectiveness of different data analytic tools/representations in the space of personal properties.