Identifying exceptional descriptions of people using topic modeling and subgroup discovery

Abstract

Descriptions of images form the backbone for many intelligent systems, assuming descriptions that randomly vary in construction and content, but where description content is homogeneous. This assumption becomes problematic being extended to descriptions of images of people [14], where people are known to show systematic biases in how they process others [19]. Therefore, this paper presents a novel approach for discovering exceptional subgroups of descriptions in which the content of those descriptions reliably differs from the general set of descriptions. We develop a novel interestingness measure for subgroup discovery appropriate for probability distributions across semantic representations. The proposed method is applied to a web-based experiment in which 500 raters describe images of 200 people. Our analysis identifies multiple exceptional subgroups and the attributes of the respective raters and images. We further discuss implications for intelligent systems.