Planning and implementing a high performance knowledge base

This paper discusses the conceptual framework for developing a rapid-prototype high-performance knowledge base for the four mission agencies of the U.S. Department of Agriculture and their university partners. These agencies include the Cooperative State Research, Education, Economic Service (CSREES), the Agricultural Research Service (ARS), the Economic Research Service (ERS), and the National Agriculture Statistical Service (NASS). The knowledge base, known as REEIS (Research, Education, Economic Information System), is a data mining application, where data are extracted from text, moved into different formats, allowing then the data mining features to run inferences, visualize connections, etc. - all generated automatically. Described are alternative data mining models along with the generalized approach to building a Warehouse architecture. Also described is the methodology used for translating system requirements into specifications and for building the REEIS prototype. The method, known as the "Rational Unified Process", is one of iteration, quality assessment, and visual modeling. The two major obstacles in the project were the normalization of disparate data repositories, and the ability to achieve an acceptable level of semantic interoperability. A metadata vocabulary model is presented to address these obstacles.