Viewfinder Case Study: Newspaper collection metadata

Abstract

The University of Kansas Student Newspaper Collection is a 189k-page digitized microfilm collection that, until this project, lacked any meaningful metadata beyond broad, reel-level date ranges, making the collection difficult to search or discover. To address this, I used open-weight generative AI models to create detailed descriptive metadata and searchable Optical Character Recognition (OCR) text for the entire collection, producing nearly 1.9 million indexed content objects and close to 300 million words, with substantially improved reading order over the library's previous OCR tool. I used Viewfinder to reflect on this project from my own perspective as the project implementer, as well as from the perspectives of the researcher, library administrator, and library IT stakeholder roles. Working through multiple stakeholder perspectives helped me see values, tensions, and assumptions that were outside of my normal considerations and helped me gain a more complete perspective on this project.