In terms of "concepts" we use the generic notion both with respect to "ontologies" (effectively "conceptual models") and with respect to "information artifacts" where concepts (we call them "concept surrogates") appear in the guise of object classes, properties, conceptual domains, and value meanings. This is a crude characterization, but it had driven much of our thinking and design.
Some readers have found this discussion quite complex. In part this reflects the desire to preserve distinctions and constructions from in ISO/IEC 11179 Edition 2, so as not to unduly burden existing users of ISO/IEC 11179 metadata registries in making the transition to Edition 3. However, our other major concern is the expectation that we will be quickly confronted with the need to manage multiple metadata registries, database schemas, ontologies, etc. We do not expect to completely conflate all of these different content sources. Instead we will have to operate in a federated environment in which many of the different content sources will continue their separate existences. Users who hope to generate successful interoperable messaging and database systems will find it necessary to create explicit mappings (linkages) among the various sorts of "concepts" in these content sources. We expect that many of these diverse "concept surrogates" will be anchored to common concepts in shared domain specific ontologies. It is this notion of federated content sources that underlies our decision to preserve the separation of "concepts" as used in ontologies and various similar notions such as object classes, properties, data element concepts, and value meanings. Instead of collapsing all of these various notions of "concept" we will maintain explicit associations, anchoring each of them a common concept in a shared ontology.
According to ISO/IEC 11179 a concept is "a unit of knowledge, specified as having a set of characteristics". Commonly concepts are used as surrogates for classes of "things" in the real world. Usually, concepts have "designators", either "symbols" (e.g., for elements in chemistry) or "names" (the more common case). "Realists" suggest that we dispense with the notion of concepts for classes of tangible "things" and simply refer to the "things". It this paper however, we follow the practice from the terminology community, e.g., ISO TC 37, and refer to all of these classes of "things" as concepts.
Concepts are commonly defined in ontologies, a.k.a. terminologies, thesauri, taxonomies, concept systems. Concepts will be connected by a variety of different semantic relationships (is-a, instance-of, ...).
Concepts will commonly be used to ground (situate in conceptual model) data element concepts, object classes, property classes, conceptual domains, and value meanings.
Finally, concepts are used (together with facets) as a means of classifying (grouping) "administered items" - i.e., the leaves (possibly instances) of a classification scheme, e.g., "administered items".
Concepts can have multiple natural language names and definitions. In the latest proposals, concepts are "designated items" in the revised metamodel, and designated items can have multiple names and natural language definitions. Concepts may also have formal definitions, which are collections of axioms (or formal statements). Formal definitions will be specified as parts of ontologies. Note that concepts may be defined by ontologies without dilineating individual formal definitions. Formal definitions will have specified "formalism", e.g., KIF, CL, OWL-DL, DAML+OIL, ...
In common parlance object classes, properties, data element concepts, conceptual domains, and value meanings would be concepts or perhaps subclasses of concepts. In the Edition 3 Metamodel for ISO/IEC 11179 we do not consider them to be concepts so we will call them "concept surrogates". The rationale for this decision reflects both some detailed technical considerations (outlined below) and more general issues generated by the anticipated multiplicity of metadata registries.
We anticipate that there will be multiple database schemas, metadata registries and ontologies. We expect that these will have to co-exist in a federated environment, in which the owner/authors of the individual database schemas, metadata registries, and ontologies will not be willing to surrender control of their respective contents to a data czar overlord. Hence, we expect that these different conceptual systems will persist and that we will be required to create and maintain mappings among them. We expect many of these mapping will be mediated by anchoring a multiplicity of various concept surrogates to a common domain ontology for particular subject areas. Typically, we expect that the anchor points will be concepts in some ontology.
There are also technical considerations which argue for linking various notions of concepts rather than conflating all these notions into one concept type.
For example, conceptual domains may have time varying membership of value meanings, so they carry time stamps for valid time domain. Conceptual domains, but not properties, are related to value meanings
If object classes, properties, conceptual domains, and value meanings were subclasses of concepts then (in common modeling usage) we would only allow one such subclass of a concept to exist. However, often we will need several of these "subclasses". Also as noted, we have multiple instances of a single "concept", e.g., multiple conceptual domains which all correspond to a single "concept". It is the potential overlap and multiple cardinality that precludes from simply specifying object classes, properties, conceptual domain, value meanings as being simple subclasses of concept.
Object classes, properties, data element concepts, conceptual domains, and value meanings are anchored to (grounded by) concepts. Since object classes and properties have no additional attributes (relative to concepts), they can be viewed as roles for concepts in the metamodel. Conceptual domains effectively specialize concepts. The concepts which conceptual domains and value meanings are linked to can serve either to differentiate conceptual domains and value meanings or to indicate that they refer to common concepts, but are differentiated either (for value meanings) by usage in distinct conceptual domains or (for conceptual domains) by having different ranges or enumerated values, e.g., the length examples cited above.
We (XMDR project) are proposing that ANSI L8 and ISO/IEC JTC1 SC32 WG2 make several changes concerning concepts in Edition 3 of ISO/IEC 11179:
This page is maintained by Frank Olken. It was last updated on 2005-09-28 at 1:18 PM EDT. (It is substantially the same as Version 7.0. of 2005-09-27 6:00 PM EDT.)