ERCIM News No.25 - April 1996 - CNR

Ontological Foundations of Knowledge Engineering

by Nicola Guarino


The word 'ontology' has recently become popular within the Artificial Intelligence community due to the growing strategic importance of knowledge integration, sharing and reuse. From this perspective, a knowledge base can in principle acquire a high added value only provided that the knowledge it contains is sufficiently independent of a specific task. This is why a rigorous ontological analysis may be important for concrete application areas like Concurrent Engineering and Enterprise Integration.

Characterised by a strong inter-disciplinary attitude, the group working on the Ontological Foundations of Knowledge Engineering at LADSEB-CNR aims to explore the synergies between Formal Ontology, Natural Language Semantics and Knowledge Representation in order to establish rigorous methodological foundations for Knowledge Engineering. Since the organization of the First International Workshop on Formal Ontology in Conceptual Analysis and Knowledge Representation in 1993, the group has played a leading role in promoting the importance of ontological issues within the Artificial Intelligence community.

The Group currently leads the CNR Special Project on 'Ontological and Linguistic Tools for Conceptual Modelling', whose goals include the unification of the existing approaches to domain analysis in the design of data bases, knowledge bases, and object-oriented systems. The Group is also among the proposers of an ESPRIT-funded Network of Excellence on Concurrent Engineering. The current research projects are the following: The problem of conceptual modelling is a crucial one for anyone who needs to use a particular representation formalism to describe an aspect of reality. The main questions include: What are the relevant primitives? How can their intended meaning be characterised? When is a given conceptual scheme better than another? Very often, the choices and the ontological assumptions underlying a particular conceptual scheme turn out to be opaque, wrong or simply incomprehensible when people change or the task at hand varies slightly, so that it becomes necessary to build a new system from scratch. The aim of the project on Ontological Tools for Conceptual Design is to make explicit such ontological assumptions by means of suitable logical theories, and to offer some methodological guidelines for the design of ontologically well-founded (and therefore reusable) knowledge bases. For this purpose, we have already done some work on the notions of Ontology and Ontological Commitment, on the various ontological distinctions among unary and binary predicates, and on the characterization of part-whole relations.

The formal representation of aspects related to space, matter, structure and function still constitutes a bottleneck for all problems related to the representation of physical entities like mechanical artifacts. The goal of the project on Logical Modelling of Mechanical Assemblies is to develop a unified logical theory accounting both for the qualitative features of simple mechanical parts (topology, dimension, form, orientation, mechanical properties of faces, edges, slots, holes...) and the possible relations among them (relative position, contact, mechanical connection, support), at different levels of granularity. Such a unified theory may play a crucial role in the integration of product data for applications in concurrent engineering and enterprise integration, since current standardisation tools like ISO-10303 (STEP) are mostly based on geometrical modelling, and a rigorous cha-racterization of qualitative features is still lacking. The relationships between mereology, topology, geometry and teleology are among the major technical issues of this project. A preliminary study has been made on the characterization of part-whole relations and on the ontological relationships between space and matter.


More information on the WWW at: http://www.ladseb.pd.cnr.it/infor/Ontology/ontology.html

Please contact:
Nicola Guarino - LADSEB-CNR
Tel. +39 49 8295751
E-mail: guarino@ladseb.pd.cnr.it
return to the contents page