Integrating Multiple Computational Techniques for Improving Image Access: Applications to Digital Collections

Museums traditionally rely on trained cataloging professionals to create metadata for their collections. While this authoritative information is well-grounded, it is brief and limited in its description of the museum objects since the human cataloging task is timeconsuming and expensive. New techniques provide an opportunity to expand subject-oriented explanatory metadata. Social tags and linguistic analysis of descriptive text holds promise, but there are many challenges to integrating these computational techniques for museum applications. In this paper, we present our initial investigations along these lines and discuss a research program to improve the integration of computational linguistics, human-computer interaction, and recommender systems to improve access to images in a museum context.