Automatic personality assessment in the wild

Abstract

Analyzing personality of a subject is an important aspect of automatic human behavior understanding. Generally, estimation of the OCEAN (Openness, Conscientiousness, Extroversion, Agreeableness, Neuroticism) traits are used to represent the personality of an individual. Personality assessment based only on individual actions is not sufficient and considering social context is also important. The focus of this Ph.D. work is to explore automatic personality assessment (APA) in the real-world environment. Most of the work till now in this area has been focusing on BF and related traits prediction of the subject in lab-controlled environments. There are several challenges involved in moving from lab-controlled environments to real-world APA scenarios such as face tracking, illumination, occlusion and social context. The early work in this Ph.D. project explores the evolution of graphs, which capture the interaction patterns and structural changes of a group of people. We call them Personality Interaction Graphs (PIG). PIGs are constructed based on the nonverbal cues to study the behavior of a subject both at a group level and an individual level. This work brings in the power of PIGs to improve the prediction accuracy and visualization of the summary of personality traits with valid cause-effect analysis at both the individual and group level. Furthermore, various machine learning techniques to analyze the personality and emotion of subjects will be explored.