Knowledge extraction from databases: Design principles of the INLEN system

Abstract

The architecture of a large-scale system for the discovery of knowledge from facts, INLEN, is described. INLEN combines database, knowledge base, and machine learning methods within a uniform user-oriented framework. Data and various forms of knowledge are managed in a uniform way by using various operators. The system's knowledge is stored as knowledge segments, which are structures that link relational tables with rules, equations and/or hierarchies. A variety of machine learning programs are incorporated into the system to serve as high-level knowledge generation operators. These operators are used for generating various kinds of knowledge about the properties and regularities existing in the data. For example, such operators can hypothesize general rules from facts, determine differences and similarities among groups of facts, propose new variables, create conceptual classifications, determine equations governing numeric variables and the conditions under which the equations apply, and determine a variety of statistical characteristics of the data. The management and discovery operators may be combined into macros or programs for repeated applications or automatic branching.

This research was done in the Artificial Intelligence Center of George Mason University. The activities of the Center are supported in part by the Defense Advanced Research Projects Agency under grant, administered by the Office of Naval Research, No. N00014-87-K-0874, and N00014-91-J-1854 and in part by the Office of Naval Research under grant No. N00014-88-K-0226, No. N00014-88-K-0397, and No. N00014-91-J-1351.