next up previous
Next: Acknowledgments. Up: Engineering Ontologies (short version) Previous: Using Explicit Ontologies in



In this article we have investigated the nature, construction and practical role of ontologies as mechanisms for knowledge sharing and reuse for some real-life industrial applications. For each of these three aspects we will summarize our main results and insights below.

Concerning the nature of ontologies, we have discussed the development of an ontology collection called PHYSSYS (Sec. 2) that covers a wide, multidisciplinary range of physical systems and their engineering. This collection contains different types similar to the distinctions proposed in [van Heijst et al., 1996]: highly generic ontologies (mereology, topology, systems theory), base ontologies valid for a whole field (e.g. technical components, physical processes, representing natural categories or viewpoints within a broad field) and domain ontologies (specializations of base ontologies to a specific domain, e.g. thermodynamics). We have indicated how we can extend this to method ontologies (Sec. 3) and how we can exploit the whole collection as the basis for application ontologies Accordingly, employing the distinctions between the mentioned types of ontologies is a natural and operational way of organizing a library of ontologies in a modular fashion.

Concerning the construction of ontologies (Secs. 2 and 3) our main conclusions are:

  1. Use and reuse of `super'theories. We have shown in detail that there are highly general `super'theories which can be employed to gradually develop large domain ontologies in a structured fashion. In our case, we have used and reused generic ontologies concerning mereology, topology and general systems theory, but it is not difficult to imagine other useful supertheories. This approach enhances both the modularity and the reusability of ontologies.
  2. Distinguishing natural `viewpoints' or base categories. In knowledge acquisition one finds that it is often possible to distinguish broad natural `viewpoints' or base categories within a field. These broad conceptual distinctions can then be exploited to develop separate base ontologies which are valid and reusable across many subdomains and tasks. In our application, these distinctions refer to groups of properties that are seen as naturally belonging together. For example, we can view an engineering system as a device configured out of known `hardware' components, or as a collection of physical processes determining its dynamic behaviour, as a thing possessing a certain three-dimensional shape, or as being composed out of different materials. Distinguishing and separating such basic viewpoints appears to be an important structuring principle in ontology building: giving rise to strong internal coherence and weak coupling.
  3. Ontology projections. We have introduced `ontology projections' as a flexible mechanism to link and configure ontologies into larger ones. There are different types of ontology projections. First, we have a technique called include-and-extend, whereby several theories are included and extended with axioms at the same level of abstraction (example: the specification of topology and general systems theory). A second technique is include-and-specialize, whereby several ontological theories are included and subsequently are specialized to a domain by instantiation, term and concept mappings and additional specific axioms (example: the process ontology). Finally, a third, new type is what we have called `include-and-project'. Here, the connection between two ontologies itself assumes the form of a full blown ontological theory. An example here is the connection between the PHYSSYS process ontology and the EngMath ontology. This example is moreover interesting because it exemplifies the reuse of an outside ontology developed by another research group in a different context.
  4. Piecemeal ontological commitment. Together, the above mechanisms provide a pragmatic approach for handling piecemeal ontological commitments in ontology development. In applications it is not so much strictly minimal ontological commitment that we want, but achieving the right commitment. This, however, needs to be built up starting from the minimal side, and step-by-step extending this by adding small additional commitments.

Concerning the practical role of ontologies, we believe that a key aspect is their capability to explicate in detail tacit background knowledge required for real-life tasks. Acquiring and analyzing this background knowledge is hard, because it is often seen as `self-evident' by domain experts and practitioners and much of it is implicitly shared by the associated community -- this is precisely why it is tacit knowledge. Bringing out this tacit knowledge is important for two reasons: (i) to find out what is really shared by the community in order to enhance reuse within this community; (ii) to develop more knowledgeable information systems that provide intelligent support for end users that are less experienced, or are from a related but different community, thus facilitating knowledge transfer between communities. In the extended version of this paper, we have given an extensive and real-life illustration of this for the domain of engineering modelling, simulation and systems design

A final note of interest is that our application has been implemented in various kinds of conventional information systems. Thus, the scope and usefulness of knowledge engineering is much wider than knowledge-based systems alone.

next up previous
Next: Acknowledgments. Up: Engineering Ontologies (short version) Previous: Using Explicit Ontologies in

Pim Borst
Fri Sep 27 13:28:43 MET DST 1996