Using Explicit Ontologies in KBS Development

G. van Heijst A.Th. Schreiber B.J. Wielinga

University of Amsterdam
Department of Social Science Informatics
Roetersstraat 15 NL-1018 WB, Amsterdam
gertjan,schreiber,wielingag@swi.psy.uva.nl

Abstract

• 1 Introduction
• 2 The Knowledge Engineering Process
  • 2.1 Knowledge Modeling Principles
  • 2.2 The Modeling Process
  • 2.3 Design
  • 2.4 Context: The GAMES-II Project
• 3 Principles for Ontology Library Construction
  • 3.1 Ontology
  • 3.2 Organization of the Library
    • 3.2.1 Issues in Library Construction
    • 3.2.2 Basic Categories of Medical Domain Knowledge
    • 3.2.3 Method- and Domain-Specific Extensions
    • 3.2.4 Structure of the Library
  • 3.3 Building the Library
    • 3.3.1 Start with an Existing Application
    • 3.3.2 Model the Application
    • 3.3.3 Scoring the Definitions
    • 3.3.4 Storing the Definitions in the Library
  • 3.4 Using the Library
  • 3.5 Summary
• 4 Model-Based Knowledge Acquisition Tools
  • 4.1 Knowledge Acquisition and Modeling
  • 4.2 A Framework for Comparing Tools
  • 4.3 Evolution of the Paradigm
    • 4.3.1 Ancient Times: Rule Editors
    • 4.3.2 Medieval Times: Task- and Method-Specific Architectures
    • 4.3.3 Modern Times: Integrated KA Workbenches
  • 4.4 Types of Tool Support
  • 4.5 Knowledge Acquisition Tools
  • 4.6 Summary
• 5 Ontology-Based Knowledge Acquisition in CUE
  • 5.1 Steps in Knowledge Modeling
  • 5.2 Skeletal Model Construction in CUE
    • 5.2.1 QUITE
    • 5.2.2 QUOTE
    • 5.2.3 Connecting Task Model and Application Ontology
  • 5.3 Model Instantiation in CUE
    • 5.3.1 QUAKE as a Passive Consistency Checker
    • 5.3.2 QUAKE as an Active Knowledge Collector
    • 5.3.3 Specialized Visualization in QUAKE
    • 5.3.4 How it Works
  • 5.4 CUE in Perspective
• 6 Knowledge-Based Integration of Representation Formalisms
  • 6.1 The Need for Multiple Representations
  • 6.2 Hybrid Knowledge Representation
  • 6.3 Knowledge-Based Integration
  • 6.4 Applying Knowledge-Based Integration
    • 6.4.1 Knowledge-Based Integration in CUE
    • 6.4.2 An Example
    • 6.4.3 Running a CUE Application
  • 6.5 Discussion
• 7 Treating Acute Radiation Syndrome: a Case Study
  • 7.1 Synopsis of the ARS Domain
  • 7.2 Modeling the Task
  • 7.3 Building the Application Ontology
    • 7.3.1 Ontology for ARS-Grading
    • 7.3.2 Ontology for Select-ARS-Treatment
  • 7.4 Extending the Library
  • 7.5 Mapping Task Model and Ontology
  • 7.6 Acquiring Domain Knowledge
    • 7.6.1 Generating the Elicitation Agenda
    • 7.6.2 Defining the Knowledge Elicitation Strategy
    • 7.6.3 Eliciting the Application Knowledge
  • 7.7 Building the Design Model
    • 7.7.1 Implementing the Problem Solving Method
    • 7.7.2 Selecting Problem Solvers
    • 7.7.3 Translating the Knowledge
  • 7.8 QUAARS in Action
• 8 Conclusions
• References

During the last decade, comprehensive knowledge-engineering methodologies have emerged which provide support for organizing the development process of knowledge-based systems. Examples are the Generic Task approach (Chandrasekaran, 1987), the Role-Limiting Methods approach (McDermott, 1988), the Components of Expertise approach (Steels, 1990), the KADS methodology (Wielinga et al., 1992) and the PROTÉGÉ framework (Musen, 1989b). These approaches share the characteristic that they promote the reuse of knowledge elements by providing libraries of off-the-shelf knowledge components. Such libraries are necessary to turn knowledge engineering from an "art" into a proper engineering discipline. So far, the emphasis has mainly been on problem-solving methods - abstract descriptions of the steps that must be taken to perform particular tasks.

Another type of knowledge which has been suggested as a candidate for reuse are ontologies - intensional descriptions of the domain knowledge in some field. Many researchers feel that access to libraries of reusable ontological components would facilitate the knowledge engineering process and several research groups have taken up the challenge of developing candidate components. However, the field is still in its infancy and many problems are unsolved or even unaddressed. To mention a few: how can ontologies be built, compared, integrated, validated, visualized or used?

In addition, the question needs to be addressed whether a methodology that is based on the use of generic problem-solving methods can also be based on the use of generic components for ontologies. In other words: can library components be specified in such a way that ontologies can be used with different problem-solving methods and vice versa?

This question touches upon a long-standing debate in AI about whether domain knowledge can be represented independently of how it is used in reasoning. Clancey's early work on NEOMYCIN suggested that both domain knowledge and problem-solving knowledge can be reused, provided that the problem-solving knowledge and domain knowledge are represented separately in the knowledge base (Clancey & Letsinger, 1984). This belief that separation of control knowledge and domain knowledge would enhance the reusability of both was also one of the assumptions that led to the conception of the KADS four-layer model (Wielinga & Breuker, 1986). However, Bylander and Chandrasekaran (1988) argued against this belief by presenting the interaction problem:

Representing knowledge for the purpose of solving some problem is strongly affected by the nature of the problem and the inference strategy to be applied to the problem (Bylander & Chandrasekaran, 1988).

The interaction problem states that the ontology of the knowledge in a KBS is strongly affected by the task of the KBS and the methods it uses to perform that task. Bylander and Chandrasekaran identified two reasons for the interaction problem. Firstly, the application task determines to a large extent which kinds of knowledge should be encoded. In general, it is not feasible nor desirable to model everything the expert knows. Secondly, the knowledge must be encoded in such a way that the inference strategy used can reason efficiently.

In this article we study the general question of how (explicit) ontologies can be obtained and used to make the knowledge-engineering process more manageable. In this context we address a number of relevant research issues. Firstly, we consider the way in which the knowledge-engineering process needs to be organized in order to make explicit ontologies useful. In order to use ontologies profitably in knowledge engineering, they must be embedded in a methodology. In Sec. 2 an overview is presented of the way in which current knowledge engineering approaches organize the KBS development process. The role of ontologies is this process is analyzed. A second issue concerns the way ontologies are obtained. Basically, there are three ways: ontologies can be constructed from scratch, they can be selected from a library of off-the-shelf ontologies, or they can be configured from off-the-shelf components. This issue is addressed in Secs. 3 and 7. Thirdly, we study the various ways in which the an ontology can be exploited to support the knowledge engineering process and to improve the quality of the resulting knowledge based system. (Secs 4-6). Finally, the relationship between ontologies and problem-solving methods needs to be studied. It is important to get a handle on the interaction problem to maximize reuse of both ontologies on problem-solving methods.

1 Introduction, 2 The Knowledge Engineering Process, 3 Principles for Ontology Library Construction, 4 Model-Based Knowledge Acquisition Tools, 5 Ontology-Based Knowledge Acquisition in CUE, 6 Knowledge-Based Integration of Representation Formalisms, 7 Treating Acute Radiation Syndrome: a Case Study, 8 Conclusions