ModernArchitectures forIntelligent Systems: Reusable Ontologies andProblem-Solving Methods

Wheninterest inintelligent systems forclinical medicine soared inthe1970s, workers inmedical informatics became particularly attracted torule-based systems.Although manysuccessful rule-based applications wereconstructed, development and maintenance oflargerulebasesremained quite problematic. Inthe1980s, anentire industry dedicated tothemarketing oftools forcreating rule-based systems roseandfell, asworkers inmedical informatics began toappreciate deeply whyknowledge acquisition andmaintenance forsuchsystems aredifficult problems. During this time period, investigators began toexplore alternative programming abstractions that could beusedtodevelop intelligent systems. The notions of"generic tasks" andofreusable problemsolving methods became extremely influential. Bythe 1990s, academic centers wereexperimenting with architectures forintelligent systems basedon two classes of reusable components: (1)domainindependent problem-solving methods-standard algorithms forautomating stereotypical tasks-and (2) domain ontologies that captured theessential concepts (andrelationships amongthose concepts) inparticular application areas.Thispaperwillhighlight how intelligent systems fordiverse tasks canbeefficiently automated using these kinds ofbuilding blocks. The creation ofdomainontologies andproblem-solving methods isthefundamental endproduct ofbasic research inmedical informatics. Consequently, these concepts needmoreattention by ourscientific community.