Tuesday, December 20, 2005
Lattice of ontologies
Function/structure descriptions
John,
The lattice notion seems attractive but has the underlying problem that I communicated in the ONTAC forum. A lattice of "theories" is needed as an interoperable mechanism (WITHOUT UNDUELY CONSTRAINING THE USER COMMUNITIES).
I feel that you have agreed that the underlying problem has to do with the imposition of too much logical constraints. (But I am never sure of your position on this point, there are practical issues in your mind and theoretical issues. We can see the conflict.)
A second potential problem has to do with the interpretation of individual terms, but this "term semantic degeneracy" (a term having more than one allowed meaning) can be managed whether by a OMG (Object Management Group) type virtual model of term meaning, or by a reconciliation of terminological differences method as seen in the SchemaLogic product. How SchemaLogic handles this is simply to have a bucket in their hash table of terms where if there are more than one meaning/context for a specific term, this is noted with a pointer in the bucket. So a single "place" does exist that specifies the precise meaning of a term, unless that term is observed to have more than one meaning/context; and then this single "location" makes this semantic degeneracy known.
One would assume that IF a lattice of logically expressed ontology were to exist in a registry, then the terminological reconciliation (ie interpretation) would occur as a first (early) step in identifying which element of the lattice was to be operational for a specific data exchange. (Right?)
Brand is suggesting we would like for this registry and lattice to work using logically specified concepts. I agree with this language "logically specified".
http://colab.cim3.net/cgi-bin/wiki.pl?CosmoWG/TopLevel
"COSMO is the proposed Common Semantic Model, viewed as consisting of a lattice of ontologies which will serve as a set of basic logically-specified concepts (classes, relations, functions, instances) with which the meanings of all terms and concepts in domain ontologies can be specified. The use of a common set of defining concepts will permit accurate interoperability of knowledge-based systems using the logical relations of their ontologies as the basis for reasoning in the system.
The exact structure of the COSMO is an issue that will be decided by the work of the COSMO working group (COSMO-WG) and its parent group, the Ontology and Taxonomy Coordinating Working Group (ONTACWG). The type and number of different logically incompatible ontological theories that may be needed to accommodate different viewpoints, and the ways they can be related to each other, will be decided by the efforts of these groups. "
<end quote>
But I note that a set of logically specified concepts does not HAVE to lead to theorem proving exercises.
We again point to the UDEF coding taxonomies (with its agreement to the ISO-11179-5 naming conventions). We also find agreement with the SUMO as well as certain ISO standards for oil and gas industry quality management.
Gary, in my review of the oil and gas terminology standards I see a constant evolution of what the industry uses. Can you make a comment about this?
The UMLS (Unified Medical Language System) is also
1) used by a significant user community (in particular by Univ of Washington's structural informatics group)
2) has become mature over the past five or six years.
I have been brought into a industry group where we will be spending some resources on the notion of a ontology hub, with a center "common set of concepts", in the spirit of SUMO's property that "all" other related ontologies make reference to the common set of core concepts.
If this common core is to be domain independent, then (I assume) that it has to
1) have only concepts that are of an abstract nature, such as the 16 classes of the UDEF and the 18 properties of the UDEF. (I am giving UDEF here only to point to a specific example of an enumeration of (what we call) a topological cover over the most general concepts needed to talk about everyday things.) I am not in favor of any metaphysical type concepts in this cover, and one finds none in the UDEF. (This is where controversy arises because, in our opinion, classical philosophy leads itself into pure abstraction, metaphysical concepts related to being, where the relationship to real things cannot be made clear.) We hope to escape the metaphysical by focusing on a structure/function interpretation of data.
2) not have a single logic imposed on the core concepts - so that specific and often inconsistent logical treatments can be added within the domain or utility ontologies.
Thus the alternative to a lattice is a hub. But in interesting cases, the hub may be the least element of a lattice of ontology with logic based theories.
But the hub may also merely be an abstract layer of concept specifications that are found commonly useful in function/structure descriptions (without a theory being imposed in the utility or domain ontology),
Comments?