(19 Nov 2024) Catalog records are key to storing and finding digital library materials. As the volume of digital materials continues to grow rapidly, the Library of Congress is exploring whether AI can help catalogers by automating the generation of metadata. AI could provide an opportunity to speed up description workflows. Yet there are numerous machine learning (ML) approaches and questions about benefits, risks, costs, and quality we must consider before adopting these technologies.
The Library recently released reports from a set of experiments called Exploring Computational Description, which examined which technologies and workflows provide the most promising support for metadata creation and cataloging, assessed the practices of other organizations, tested many different ML approaches with Library ebook data, and evaluated the output in iterative data review interfaces.
Leah Weinryb-Grohsgal, Senior Program Advisor to the Director of Digital Strategy, recently interviewed Abigail Potter, Senior Innovation Specialist in the Library’s Digital Innovation Division (LC Labs) and Caroline Saccucci, Chief of the U.S. Programs, Law and Literature Division, about their hopes for this experiment. The group also discuss how they’re interpreting the automated outputs and user implications of adopting AI for a core library workflow.
Find out more here.
