What Can AI Do for Special Collections?

Abstract

ABSTRACT This research project investigates the question, what can AI do for special collections by testing thirty-three software tools on a manuscript collection. The authors provide an in-depth analysis of their experiences with specific AI tools applied to untranscribed documents (handwritten and type-written) and photographs in the William Elliot Griffis Collection at Rutgers University Libraries. The evaluated tools range from end-user generative AI to APIs requiring Python programming and include software that is commercial and open-source; fee-based and free; web browser-based and standalone; and run in desktop, mobile device, and high performance computing clusters environments. Application scenarios extract text from images, perform complex pattern matching, and generate metadata. Each scenario discusses audiences, ethical considerations, tool evaluations, and implementation suggestions. Collaboration among cultural heritage professionals, humanities scholars, and computer scientists addressed challenges of resources, communication, and policy. The project successfully demonstrates the potential of AI to improve accessibility and discoverability in manuscript collections.

Ähnliche Arbeiten