To appear in CACM (Communications of the Association for Computing Machinery) April 1995 special issue on Digital Libraries
The University of Illinois is building a large-scale digital library testbed, planned to grow to thousands of users and thousands of documents, with the goal of bringing professional quality search and display to Internet information services. Concurrently, research is designing and implementing a prototype of the Interspace, a vision of what the Internet will evolve into, where the distributed network of interconnected machines is replaced by a distributed space of interlinked information.
The testbed collection consists of articles from engineering and science journals and magazines, obtained in SGML format directly from major partners in the publishing industry. This collection will be managed by the University Library on a production basis, growing into a standard service of the new Grainger Engineering Library Information Center.
The testbed software will support comprehensive search and display of complete contents of articles, including text, figures, equations, and tables. The software is based on NCSA Mosaic as a multi-platform World-Wide Web connection to commercial software, currently SoftQuad Panorama for SGML display and Dataware BRS for fulltext search.
The National Center for Supercomputing Applications is developing a custom version of their Mosaic for this testbed with sufficient client and server interfaces and gateways (e.g. CCI and Z39.50) to bring professional display and search to widely deployed Internet information services.
The testbed users will be faculty and students at the University of Illinois initially, then spread to the CIC consortium (Big Ten universities).
The user evaluation will interview hundreds of users in focus groups to provide detailed cognitive descriptions of needs and uses, plus survey thousands for a grosser statistical picture. The software will also be instrumented to learn how to deduce more detailed ethnographic information from large-scale network usage.
The technology research efforts are centered around scale and functionality, and will migrate into the testbed as they prove effective. Providing semantic retrieval at a deeper level than commercial search is necessary to support wide ranges of users across wide ranges of collections.
The utility of physical library classifications for networked digital collections is being investigated by experimenting with interfaces using major classification schemes (e.g. Dewey Decimal, Library Congress, INSPEC thesaurus). A complementary effort is to generate classifications automatically. This will be tested using a concept space approach based on co-occurrence matrices, which has proven effective in specific domains such as molecular biology.
The research efforts will be assembled into a next generation system based on a new architecture for the Interspace. This will provide an environment for an information space of structured objects across the network. Such an environment will enable information sources and services (data and programs) to be plugged into the space, while supporting interactive functionality. For example, a user could execute an equation from an article while browsing or record a navigation path through the collection to share with others. Support of analysis and communication for programs and people will provide a new level of functionality for network information systems.
PIs:
Bruce Schatz, Ann Bishop, Bill Mischo, Tim Cole, Joseph Hardin, Larry Jackson Graduate School of Library & Information Science, University Library, National Center for Supercomputing Applications, University of Illinois, Urbana-Champaign, IL 61801, thabing@uiuc.edu
Partners:
IEEE Computer Society, American Physical Society, American Institute Aeronautics& Astronautics, American Society Civil Engineering, Institute of Physics, John Wiley & Sons, Tribune Company, SoftQuad, Dataware, Spyglass, United Technologies, Corp National Research Initiatives, Univ Arizona, Carnegie-Mellon University.