Open Competition 2 - Information Technology

Automated Knowledge Discovery System (AKDS)

Sponsor: InRAD, L.L.C.

• Project Performance Period: 8/1/2003 - 1/31/2006
• Total project (est.): $2,372,029.00
• Requested ATP funds: $1,996,181.00

Faced with a sea of data and sources, planners and researchers in all fields are increasingly challenged to locate, organize, and analyze knowledge that is relevant to their interests. InRAD proposes to develop the Automated Knowledge Discovery System (AKDS), an automated software system designed to search for and organize content from Internet sites and databases that precisely matches a user's information requirements. These requirements may be drawn from any narrative describing the research and development or subject of interest, such as an organization s strategic plan. Once these information requirements are entered into AKDS, the system will automatically generate a domain-specific ontology (model of key vocabulary words and the semantic relationships among them) to direct the search for relevant document content, which AKDS then evaluates and classifies. AKDS will update collected information continuously and operate with high speed, precision, accuracy, and minimal redundancies and inconsistencies. This two-year project relies on two vital subcontractors. Knowledge Based Systems (College Station, TX) will develop the AKDS toolset subsystem, which combines machine-learning techniques (text and data mining) with domain ontologies to enable ontology-assisted knowledge extraction from a variety of information sources. Sarnoff Corporation (Princeton, NJ) will develop the search subsystem, which performs targeted searches in domain-specific databases and trusted sites on the Internet for knowledge related to specific user requirements. This project presents high technical risk because integrating ontologies with machine-learning techniques has never been attempted. InRAD has been unable to attract financing from venture-capital and incubator organizations, which found that insufficient progress toward a prototype makes the project difficult to assess and thus too risky. Without ATP funding, this project cannot proceed. If successfully developed, AKDS could speedily glean and organize substantial quantities of relevant content from many available sources, avoid duplication of research efforts, and help an organization better direct its R&D. In 2002, U.S. industry and government spent an estimated $285 billion on R&D. Through effective management of research and development resources across industry, research institutions, academia, and government, the savings generated by AKDS in time and labor costs could amount to billions of dollars.

For project information:

Richard Neal, (865) 947-7000
imti1@msn.com

ATP Project Manager

Barbara Cuthill, (301) 975-3273
barbara.cuthill@nist.gov