Journal of Systems Integration, 10, 375±394, 2001 # 2001 Kluwer Academic Publishers. Manufactured in The Netherlands. Parallel Automated Knowledge Acquisition of Case-Based Semantic Networks from Relational Databases AUBREY E. HILL hill@cirrus.biosccc.uab.edu Comprehensive Cancer Center, University of Alabama at Birmingham WARREN T. JONES jones@cis.uab.edu Department of Computer and Information Sciences, University of Alabama at Birmingham (UAB) ROBERT M. HYATT hyatt@cis.uab.edu Department of Computer and Information Sciences, University of Alabama at Birmingham (UAB) J. MICHAEL HARDIN mhardin@uab.edu Department of Health Services Administration, School of Health Related Professions Abstract. The number of databases that are accessible over networks within organizations is increasing. This paper presents a methodology for automatically converting the data in these databases into a useful knowledge base of case-based semantic networks that can be accessed through a browsing facility. A parallel processing strategy has been implemented for this knowledge acquisition process to support its scalability to large databases. This methodology has potential application in the development of organizational intranets. It can also be used for retrospective browsing of the context of interesting patterns discovered by data mining. The database examples used in this paper are from clinical laboratories that provide data to a hospital infection control committee. Even though the results presented here use a single domain, the methodology can be used with no changes to explore the construction of multidomain knowledge bases. Keywords: knowledge acquisition, case-based reasoning, relational databases, semantic networks, parallel processing 1. Introduction The capability of accessing enterprise-wide knowledge is becoming increasingly important to the competitiveness and effectiveness of organizations. This enterprise- wide perspective is possible only if numerous diverse databases of information can be transformed into a knowledge representation that is accessible in a useful interactive mode. This paper presents a methodology for automating this knowledge acquisition from existing diverse relational databases. This methodology has potential application for the development of organizational intranets. It can also be useful as an approach to retrospective browsing of the context of interesting patterns discovered by data mining techniques [1]. Similar approaches to knowledge base construction have been explored by others [2±5]. These approaches have included either substantial manual construction components and/or goals of using the acquired knowledge for prediction. Our approach automates the construction process and uses case-base reasoning technology [6] as the primary basis for organizing and browsing the knowledge.