Artificial Intelligence for Content and Context Metadata Retrieval in Photographs and Image Groups
Peter Fornaro, Vera Chiquet · Archiving Conference · 2020
Cataloguing is a frequent bottleneck in the digitisation of analogue images as it is impossible to scale the necessary content-related knowledge. As this paper discusses, however, it is possible to use well-trained artificial intelligence to semi-automate metadata enhancement for photographic collections. The paper describes a study in which participants indexed historical collections of photographs. In a subsequent interdisciplinary project with contributions from cultural anthropologists, computer scientists, digital humanities researchers and art historians, these descriptions and indexes were then used to train machine-learning components. With this interdisciplinary approach, it is hoped that cataloguing practices can be enhanced, generating new insights into AI and semantic metadata.