Wednesday, November 23, 2005
Center of Excellence Proposal
à
ONTAC stands for Ontology and Taxonomy Coordinating Working Group
It is a working group of
Semantic Interoperability Community of Practice (SICoP)
Discussion Subject: Re: [ontac-forum] Hub ontology
Communication from ONTAC from Cory Casanave
All,
Within the GSA sponsored "OsEra" project we are currently engaged in creating a "mid" ontology with many goals and attributes in common with those expressed here - called "Semantic Core". Rick Murphy (who is also involved in this same project) suggested that we should enter the discussion.
The problem space we are addressing is the fragmentation of formal and informal architectures. Architectures include business and technical views, process, information, finance, goals and objectives, plans and specifications.
This space spans the range from "Enterprise Architecture", domain Ontologies and business modeling to systems architectures, implementation architectures and, of course, architectures of interactions (E.G. information flow). Most architectures are a mix of context, description and specification - however the purpose of specification is central to the reason most architectures are done.
We characterize this as the specification of systems where "system" includes any organized construct including any business, organization, community or computer program. The problem is that architectures are done with a wide variety of languages, formalisms, vocabularies and styles. The specification of a single system (or system of systems) ordinarily involves multiple formalisms (languages, notations, theories, etc) that are not well coupled and yet represent overlapping concepts. This is a source for error, confusion and redundancy. There are a set of problems dealing with the domain concepts in these architectures - (E.G. do you and I mean the same thing by "purchase"). There are also a set of problems dealing with the concepts of doing architecture (E.G. do you and I mean the same thing by "process"). We are dealing with the later set of problems, these are essentially the "meta concepts" common across architectural languages and notations.
Languages typically used for these purposes include UML, E-R models, OWL, Collaboration Modeling, Services Interfaces, Information Models, FEA-RMO, etc.
The approach is to normalize and unify the concepts expressed in these various languages into a controlled but open set of concepts, this is the "semantic core".
These concepts may be introduced from any of the architectural languages - our job is to try and "slice and dice" the concepts so that the fit together (where possible) and are non-redundant (where possible).
We can then describe the mapping and/or transformation of various tools and representation into this common form. Thus it is certainly a "hub" as described within a specific domain (of architectural description), but the domain is quite wide. This will have value in fully specifying an architecture as well as unifying architectures across time, focus and organization.
This probably puts the work at the upper end of "mid" or "reference" Ontologies but certainly does not require kind of upper ontology required to understand human knowledge in general. We refer to it as an ontology (or meta model) of architecture. To try and solve this without some kind of hub would suffer the "N Squared" problems but would also suffer the problem that complementary information about the same thing would not be represented in any one place. (from BCNGroup: see comment on categorical collapse à [234] )
In keeping with the unification of concepts across representations, this set of concepts is being represented as both an OWL ontology and a UML/MOF meta model.
We are also currently automating the structural transformation of object oriented meta models (this includes most software engineering disciplines - such as UML) to OWL. Note that this work actually started more in the structural modeling "camp", using the "Meta object facility", UML, etc. But we feel that there would be a lot to gain from the capabilities offered by ontology and semantic web techniques and tools.
While we very much believe (and are along the way to demonstrating) that unification at this level is possible, practical and useful, we do not suggest it is "the one and only" hub - this is perhaps the dangerous idea.
However, given success and use, it could become a hub in common across a substantial number of representations and architectures and thus provide substantial value as well as a place to join with other hubs. Semantic core it's self is not currently grounded in any upper ontology but I could see it possible and useful to do so, and perhaps (where possible) ground concepts in multiple hierarchies. This fits with John Sowa's approach for modules. This work would both be a module as well as be composed of modules.
In that any "hub" gains value by a larger community it certainly makes sense to us that ontac be involved with the care & feeding of hubs - such that we not have to few or to many. It would be interesting to compare notes with some of the other efforts mentioned to see where the hubs fit together. The status of the semantic core is still early.
Cory Casanave
Data Access Technologies GSA,
OsEra Project
www.SemanticCore.org