Automated Visual Clustering: A Technique for Image Corpus Exploration and Annotation Cost Reduction

Description

Compared to text and voice, images can be an especially effective form of political communication. It is now relatively easy to automatically label images for many categories of interest (such as protests, famous people or facial expressions). As a result, scholars are increasingly using large-N image analysis to investigate contemporary political attitudes and behavior. We address two emerging needs of image scholarship. The first is that researchers may want to visually explore an image corpus to discern patterns before they begin assigning labels. This can be difficult with very large corpora. The second is that researchers may be interested in image categories that cannot be easily assigned using off the shelf automated methods. We demonstrate how unsupervised image clustering can help researchers to address both of these needs more efficiently. Our substantive focus is on exploring and labeling a large corpus of images shared by Twitter users along with the hashtag #FamiliesBelongTogether.

Period	13 Oct 2020
Held at	University of St. Gallen, Switzerland
Degree of Recognition	International