Integrating An Image Database into the World Wide Web

Project objectives are to:

1. Establish a World Wide Web image database and demonstrate a classification and retrieval system that enhances the World Wide Web search paradigm by creating structures that cluster art images into browsing categories.

2. Test the effectiveness of the classification and retrieval design by demonstrating this system to subjects having information needs associated with the images of the history of art. The testing will include an interactive process by which the informat ion users participate in the design

3. Apply the results of the testing to the implementation of the classification and retrieval system thereby extending accessibility of World Wide Web to any user interested in history of art images.

4. Disseminate the research findings through publications in the professional literature. An interdisciplinary group consisting of members of the user community (both general users and domain specialists), classification and retrieval specialists, evaluation specialists, and computer scientists will collaborate on the project from its inceptio n. A planned process of design, prototype implementation, and user evaluation will ensure that the development of the system will be interactive and iterative rather than strictly sequential.

In developing classification structures for browsing, the project team assembled at Michigan will question user populations to determine information needs, patterns of use, and information searching practices. The responses of the users will play a large role in the emerging classification scheme. As consultant members of the project team, faculty members in the University's History of Art department and at the Museum of Art, student users, librarians, and other specialists will be asked to evaluate the r esponses and to work with the classification specialists in categorizing the art images. By using the analysis of the user consultants and drawing upon subject classification models, the project team will construct broad subject outlines. The outlines wil l be modified and refined through tests of the categories by the specialists and users with a need for information contained in the art images.

These refinements will be integrated into classification structures which cluster images into browsing categories. The concept of browsing is integral to this project. Browsing relieves the user from the burden of formulating a precise search strategy and capitalizes on the fact that it is easier to recognize what is interesting or useful than it is to specify it in advance. A related advantage is the ability to facilitate exploration without prior knowledge of subject content, thus enabling users to prof itably cross over into unfamiliar domains. Furthermore, browsing lends itself particularly well to visual images since, unlike text, the graphic image is able to convey its message all at once and can describe itself in its own terms to the viewer. Previo us methods for accessing image databases have relied primarily on either indexing of individual items or automatic image recognition. The present research emphasizes classification structures as devices to group image sets into meaningful categories that support browsing.

In addition, a search engine called The Full Text Lexicographer (FTL), a locally-produced retrieval application, will provide enhanced search capabilities (keyword and Boolean searching of record fields and full text describing the images).

The management of visual information is emerging as a key challenge in the development of information technology systems. The proposed project presents one possible strategy for system development, and uses the concrete example of an existing database in art history. However, the proposed strategy is not limited to art history, but can serve as a model for other image databases. The proposed general purpose image facility can have wide application to the management of image information in a diverse spectr um of disciplines.