| .................. | DLI-2 PROJECT SUMMARY The Net will dominate infrastructure in the 21st century, with the ubiquity of personal computers and information sources nationwide. Healthcare will dominate the economy in the 21st century, since it is already 15% of the GNP in an aging population. To ensure quality in the era of managed care, information infrastructure must support less trained professionals. Analysis environments to correlate information across sources would provide such support. This research project will develop and deploy an experimental testbed of the future Net of information infrastructure and the future of information analysis in providing healthcare. The Interspace is a vision of the Future Net where every small community maintains its own repository of its own knowledge. People find items in the Interspace by traversing spaces to interconnect knowledge. Handling many small sources will require that the classifications must be deeper and the correlations must be stronger than at present. The information infrastructure must thus provide substantial support for semantic indexing and semantic retrieval. For Digital Libraries, the Grand Challenge [IITA] is semantic interoperability across heterogeneous collections. For text collections, semantic interoperability is vocabulary switching across subject domains. Our approach is concept switching across semantic indexes on community repositories. We expect to demonstrate practical large-scale semantic interoperability for discipline-scale collections during the course of the proposal period. The MEDSPACE project is about implementing the Interspace for medicine and biology, by developing research technology for semantic federation of community repositories, and deploying that technology in a large-scale testbed for clinical medicine. The sources will be medical literature, both bibliographic abstracts and online full-text. The users will be practicing physicians and biomedical researchers at different levels, from rural clinics to urban hospitals. Large-scale semantic indexes on bibliographic databases will enable a simulation of the world of community repositories, by partitioning these medical databases into community-sized collections. Nearly a billion objects will be indexed by partitioning MEDLINE and BIOSIS into nearly a hundred thousand community repositories, and computing concept spaces using allocations on high-end NCSA supercomputers. These semantic indexes will be used for term suggestion on the bibliographic databases and full-text sources in clinical medicine. The system will be field tested in the Department of Surgery at the University of Illinois at Chicago (UIC), followed by the UIC Medical Center, a high-volume urban site, and the Library of the Health Sciences. We will deploy prototypes throughout the HealthAlliance HMO, which dominates rural Illinois, to complete a state-wide testbed. We will deploy successive versions of MEDSPACE, reaching first tens then hundreds then thousands of potential clinical users. Longer-term sociology research will evaluate the usage of the testbed, plus investigate novel training techniques and useful metrics for evaluation of new digital library systems. Longer-term technology research will develop concept switching across automatically classified collections, dynamic indexing via generating custom collections during a search session, and user modeling to guide search via custom profiles for situational analysis specialized to user and to task. This research project is focused at the heart of the information systems needed for the technophile but aging baby boomers. It is based in the heartland of Illinois with a rural and urban mix, where the problems and solutions of the next century will be encountered first. MEDSPACE will be the model for providing electronic medical information for the Interspace. The technology will be a model for the information infrastructure of the 21st century, when information analysis is a routine operation and ordinary people do problem solving on the Net. |