Bottom-up construction of ontologies: the case of an ontology of pure substances

Bottom-up construction of ontologies departs from the more usual taxonomic approach to ontology development. It allows for parsimonious speci cation of concepts. De ned concepts are written as compact expressions that allow exible reasoning. Obvious applications include chemistry and engineering design. For such domains, the method enables better ontologies than the taxonomic method. Taxonomic approaches remain appropriate for other applications. Bottom-up ontology construction is illustrated by presenting a drawn-out example involving an ontology of pure substances. This ontology is a simpli ed version of a part of the ontology developed for the Plinius project of the Knowledge-Based Systems Group, University of Twente. The viability of the approach is demonstrated by presenting a complete formalisation in pure logic programming. Chapter 1 Foreword In the Plinius project of the Knowledge-Based Systems Group at the University of Twente, ontology design is an actively pursued research subject. In 1991, we proposed an ontology for the domain covered by Plinius in a technical report and several publications. Since then, many discussions and a closer look at our needs in the project necessitated a large number of revisions. In the course of those discussions we also realised more clearly on which principles the Plinius ontologies are built. Of these, the present report discusses two. The rst issue is the function of an ontology in the whole process which ultimately results in an implemented knowledge base. The second issue is a particular way of building ontologies called the bottom-up method. The report may thus seem to address one issue too many. As the title makes clear, the bottom-up method is the proper subject. However, we found that important aspects of the discussion are left hanging in mid-air if we omit a discussion of the rst issue. For the purposes of discussing bottom-up construction of ontologies, we have taken the subject of pure substances out of the larger ontology. Pure substances are su ciently complex to merit discussion, but su ciently simple to serve as an illustration. Also, the subject of pure substances takes relatively little domain knowledge. This report discusses two variants of the bottom-up principle. In the simpler variant wholes are simply de ned by telling which constituents there are. The other variant in addition speci es how the constituents are connected. Judging by the examples of non-chemical applications presented in chapter 6, it may seem that the applicability of the simpler variant is limited. The fact is, we don't know. Further research will have to clarify these matters, but as there is at least one important application we did not want to wait. In preparing this report, we have pro ted from the many discussions with the other members of the Plinius/Condorcet research group: Piet-Hein Speel (now at Unilever Research Laboratories, Vlaardingen, the Netherlands), Wilco ter Stal, Jeroen Nijhuis, and in particular Hidde de Jong. We are also indebted to Stef Joosten (University of Twente) and to Tom Gruber (then at Stanford) for their comments on various versions of the ontology. Enschede, September 20, 1995. Paul E. van der Vet Nicolaas J.I. Mars Address of the authors: Knowledge-Based Systems Group Dept. of Computer Science, University of Twente P.O. Box 217, 7500 AE Enschede The Netherlands 1 Foreword 2 Phone +31 53 89 36 90 Fax +31 53 33 96 05 Email fvet,[email protected]