Presenter
Natalie Milbrodt (@NatalieMilbrodt)
Bridget Day (@br1dg3t.day)
Richard Knipel (@Knipel)
Abstract
In this presentation by the City University of New York’s University Archivist, Wikimedian-in-Residence, and Digital Archivist, they’ll share how they followed their values and navigated through their AI apprehension to make a local pantheon of “CUNY-famous” individuals more discoverable on the web. They did this work with the help of a volunteer-built tool called the “CUNY BoT Bot” that scanned every page of CUNY’s publicly accessible 84 years of Board of Trustees meeting minutes for personal names and then checked against Library of Congress name authorities, Wikidata entity records, and VIAF authorities for those individuals. The tool incorporates AI to extract text from PDFs of the board minutes, deduplicates name variations across many pages within a PDF file, and synthesizes a brief biographical note for certain flagged authorities (based only on the corpus) in a non-agentic scripted data pipeline, while the tool itself (both pipeline and visualization UI) was developed with AI coding assist tools. Details about the CUNY BoT Bot development process that implemented AI coding assist tools are available on the About page of the tool’s website.
The team is exploring further development of a generalized, web-based tool for use with any set of digitized documents, based on community interest. In the meantime, CUNY archivists continue to use BoT Bot to help them identify the “CUNY famous” still in need of an authority record. By keeping humans in the loop for important analytical tasks and not designing BoT Bot to flood Wikidata with AI generated entity records of questionable accuracy, the team feels they have navigated an ethical and efficient use of AI to improve online discovery for their collections while ensuring the authenticity and integrity of their metadata.