James Shulman
President
ARTstor
William W. Ying
Chief Information Officer
ARTstor
Finding useful search results is an ongoing challenge to users of large-scale digital libraries where a typical keyword search result can return thousands of records. Resource providers can help users to sift through Web-delivered content and build better result sets by drawing upon the collective preferences of other users. ARTstor recently released its Associated Images feature, which anonymously mines user preference data via collaborative filtering. As users at over 1,100 institutions make image groups for their own purposes, they are also revealing preferences with each image that they include. Just as Amazon “recommends” other books that are often purchased by the same people who purchased a given book, the collective choices of scholars who have compiled ARTstor image groups are now shared with other users across the ARTstor network.
There are many ways to enhance search, ranging from the difficult work of metadata enhancement to personalization to filtering results by various criteria. The session will report on the choices made in the preparation and release of this feature and is intended to provoke discussion about other methods that can also be employed to provide users with particularly meaningful results as they navigate large complex datasets.
http://www.artstor.org/news/n-html/an-090205-tip.shtml
Handout (PDF)