Reimagining Audio-Visual Accessibility and Discoverability with AI Tools

The University of Utah J. Willard Marriott Library’s Special Collections department has had an expansive approach to acquiring audio-visual materials throughout its 50-year history, but many collections remain difficult to access because of format obsolescence and processing challenges inherent in time-based media. Now, an increase in format specialists across departments and emerging AI technologies have presented an opportunity to enact new plans for discoverability and accessibility alongside needed digital preservation. The Marriott Digital Library team chose to focus on the most time intensive audio-visual materials for its initial experiments with AI. The urgency for the need to digitize and digitally preserve media at the brink of obsolescence made it a priority, and the unique descriptive metadata needs of the KUTV broadcast news collection an ideal candidate for experimenting with AI-assisted workflows. The Marriott Digital Library team decided on utilizing AI assistance in automated practices while pursuing a human-in-the-loop model to prioritize rich, accurate metadata for materials in the digital library. For Special Collections, the increased accessibility of digitized audio-visual materials creates new avenues for researchers to discover unique primary source media. Already, human and AI-assisted efforts to improve discoverability through detailed description have boosted requests for audio-visual collections. With the implementation of new AI-assisted accessibility functionality such as searchable, timecoded transcriptions, we expect this trend of expanded interactions and requests to continue as previously hidden collections become more discoverable and usable for a wider range of researchers and community audiences.