Description of Results, Products, Transferable Technology, and Technology Transfer Path
Products and Results
Algorithm packages
Each of these packages will have software for indexing a collection and a Web interface to support semantic retrieval for the experimental collections. User evaluation experiments will be performed on the large-scale collections and published papers available describing the algorithm techniques and the evaluation results. Real-time implementations of the algorithms will be available in later years of the project.
Algorithm packages include those for:
concept spaces (on both text and image collections),
category maps (on both formal text like journal articles and informal
text like email messages),
meta-data generation (for users to index their own collections), and
meta-map (to support vocabulary switching across concept spaces).
Analysis environments
The major product will be the Interspace prototype which is a complete analysis environment that supports semantic interoperability (across media and across subjects). The prototype part is a software package suitable for distribution to the DARPA community. The Interspace part is a set of large-scale test collections in various subject domains of interest to the DARPA community (engineering, science, medicine). The semantic interoperability supports analysis correlations at different levels of abstraction (categories, concepts, objects) across all collections. The functionality of this environment will be described in detail in an architecture document, using the software methodology from the Design Patterns community. In describing the architectural frameworks for analysis environments, we will be collaborating with the UI group led by Ralph Johnson, one of the worldÌs leaders in design documentation for software environments. Various technical papers about the design and implementation and usage of the system are also planned.
Transferrable Technology
A typical technology transfer would be for a researcher to obtain the prototype client and perform semantic analysis on the existing collections across the Internet. The client requires Smalltalk for display and CORBA for communication; it runs on Unix workstations and Windows PCs. The repositories would reside in this case at our central site on existing NCSA servers.
For researchers wishing to maintain their own repositories generated from their own collections, facilities will be available for semantic indexing. These software programs and processes will generate indexes for the different levels and create user repositories from either formal (e.g. technical documents) or informal (e.g. email messages) local collections. The user repositories can be then incorporated into the Interspace as fully-fledged portions. Semantic analysis across distributed repositories, physically both local at the user site and remote at our central site, will then be available.
Technology Transfer Path
We expect to have sufficient external users and repositories to demonstrate a national model for transforming the Internet into the Interspace. This includes ranges across media, subjects, sites, and levels. As a start, this will include both our main sites at the University of Illinois (MidWest) and the University of Arizona (SouthWest). It will next expand to our close technology collaborators (Ralph Johnson on software engineering at Illinois and Jay Nunamaker on collaboration infrastructure at Arizona) and our close applications collaborators (University of California at Santa Barbara for map image repositories and National Cancer Institute at NIH for medical concept mapping).
Finally, it will include many of the sites from the new NCSA (National Computational Science Alliance) which are distributed in major technology centers across the United States. The new NCSA is part of the new NSF PACI (Partners for Advanced Computational Infrastructure) program. Our main sites are technology partners in NCSA PACI -- the Schatz lab is the partner for digital libraries and the Chen lab is the partner for information science.
The technology transfer path will expand as the software environment becomes more mature and the range of collections broadens. We expect that Year 1 will be primarily experiments to solidify the algorithms, Year 2 will be primarily internal usage of the environment on small collections, and Year 3 will be primarily internal usage on large collections. The end of the grant period will see a complete analysis environment available for distribution to the DARPA community.
4. DESCRIPTION OF
RESULTS,
PRODUCTS, TRANSFERABLE TECHNOLOGY,
AND TECHNOLOGY TRANSFER PATH
7. TECHNICAL RATIONALE, APPROACH, AND PLAN
8. COMPARISONS TO OTHER RESEARCH
