Research: Media Management

In addition to my work on digitization of historical materials, I have also worked extensively in large-scale media management, developing intelligent systems that observe human behavior and automatically reorganize image collections, as well as automated licensing systems that oversee the licensing operations of the University of Illinois' premier image archives.

We live in a media-rich world where every cellphone is a camera and visitors to Facebook alone upload more than two billion photographs each month. Digitization and digital history have seen explosive growth in recent years with the advent of bulk-digitization workflows like Google Books and the Open Content Alliance. Current media management platforms have not kept up with this explosive growth: either encumbering users with outdated interface modalities, or siloing information by each aspect of a user's life.

Automated Ranking

All large-scale image collections run into the same problem: how to rank images according to individual user queries? If there are 5,000 images that all match the request search term, how can a system rank those images to respond with the most relevant ones first and how does the definition of "relevant" vary by user? Using the living laboratory of the UIHistories Photographic Archive over the last half-decade, I've developed a number of approaches to the automated ranking problem that are in use on the production site.

Automated Licensing

Large image collections that are in high use require continual licensing of images. Each licensing grant requires evaluating the request use against the available uses of the image, checking rights clearances, appending additional use-specific specialized limitations on the license as necessary, and logging and notifying copyright holders of the license. A fully automated licensing system was developed for the UIHistories Photographic Archives and has been in use for five years, overseeing more than 22,000 licenses.