• forward to Theory of Meaning in ONIONS
• forward to ONIONS Methodology
• forward to ON8.5 Top-level Description
• forward to Ontology of Meta-level Categories

Geri Steve, Aldo Gangemi

Reparto Informatica Medica, Istituto Tecnologie Biomediche, CNR

INTRODUCTION

We introduce ONIONS (ONtologic Integration Of Naïve Sources), a methodology for coping with the knowledge integration problem. We describe the ONIONS theory of meaning, its procedure of knowledge integration, and some results of its application to the medical domain, including a description of the top model of ON8.5 ontology with an explanation of its commitment.

ONIONS is based on the belief that many ontologic problems of knowledge acquisition come from the need for sources of motivated knowledge, which are eventually heterogeneous in their conceptualizations. Once such sources are taken into account, immediately the knowledge integration problem arises. ONIONS treats it in an explicit and rational way, i.e., by assuming that relevant knowledge in a domain could be extracted by understanding and generalizing the conceptualizations inherent in the lexical repositories of a domain.

On this premise, we propose a layered approach to ontological engineering:

• theory of meaning: some well-suited background assumptions about the relation among world, language, and thought. A theory of meaning should prevent from confusion among the categories concerning the aspects of knowledge [cf. Gangemi, 1995b], and should provide a milestone to ontological commitment;
• ontology acquisition: methods for acquiring a motivated semantic field within the conceptualization of a domain. An ontology acquisition method should produce a so-called non-formal ontology, say, an explicit, schematic account of the conceptualization of a domain. An account of what we mean by conceptualization and semantic field is detailed in par. 1.1;
• concept structuring: methods to formalize concepts from a semantic field by a set of meta-level categories (henceforth, MLC). This should provide a partial semantic account of the intended conceptualization of a logical theory;
• knowledge representation: some well-suited formal language for modelling a semantic field. Such a language should allow to generate a formal ontology, say, a logical theory which gives an explicit, partial account of a conceptualization.

Such definitions are partially derived from several discussions about differences in the intended meaning of ontology [Cocchiarella, 1991; Gruber, 1993; Guarino, 1995; Gangemi&al, 1996]. The following is a further schematization:

In [Gangemi&al, 1995a; Steve&al, 1996; Gangemi&al, 1996] we introduced ONIONS, a methodology for aiding the creation of a domain formal ontology from the integration of distinct, authoritative taxonomic knowledge repositories. ONIONS allows to perform ontological analyses with a "built-in" solution to the knowledge integration problem. The integration approach of ONIONS is suited to naturally introduce an ontological commitment for the theories derived through its application.

The main application of ONIONS has been made in the domain of medicine. The resulting medical ontology is currently ON8.5 (in [Gangemi&al, 1996] an appendix is presented which contains the formal model of ON8.5 in an order-sorted-logic.

In par. 1. we describe the theory of meaning underlying ONIONS; in par. 2., the ontology acquisition methodology; in par. 3., some description of ON8.5 top-level; in par. 4., some concept structuring issues: the decision tools for tagging concepts by MLC, and some mapping of our MLC to the ontology of MLC introduced by Guarino [Guarino, 1992; Guarino&al, 1994; Guarino&Boldrin, 1993].