Automate it: A Deep Learning Solution for Library Archives




England, Cale
Prud'homme, Patrice-Andre
Soliday, Hunter

Journal Title

Journal ISSN

Volume Title


Texas Digital Library


In the fall of 2018, the Oklahoma State University Library started to look into machine learning to increase the visibility of archival collections. The primary focus of this project was to create automata that would assist in inventory work, focusing on metadata. This project created a methodology that, using universal policy, is able to incorporate all forms of metadata to a single format to address inconsistencies in existing metadata. Using this framework we began adding general metadata tags (“dog”, ”cow”, etc.) created via deep learning. The second phase of this project was a facial recognition databasing system that was designed with the intent of being able to trace individuals featured in the Oklahoma State University yearbook collections to works in other archives, in order to give said works a more general context. This presentation will focus on the techniques used for this project, explained for the purposes of being used in other digital collections.


Presented by Oklahoma State University, 1C | Technology & Tools, at TCDL 2019.