Sunday, December 11, 2005
(new thread on Emergency Medical Ontology Project planning ŕ [home] )
In reference tohttp://www.mitre.org/work/tech_papers/tech_papers_04/04_0603/04_1175.pdf
I point out that the report's conclusion is that there may exist no single correct upper ontology. I agree, however there is a test of sufficiency. SUMO is sufficient, clear and is not (overly) entangled with first order logics or proprietary software. It is merely a set of concepts, relationships, attributes, and subsumption relationships. I would flatten the set of concepts so that there were no subsumption relationships, but I am allowed to do that without harming it.
Why I might consider flattening the subsumption relationships is a different story - one related to a physical model of how human mental events form. (A mental event is a single thing, with entailments.) I am also able to separate the hooks for standard logical inference and treat these situationally. (My group is working on an elaboration of what we mean, but the simplest explanation is that a concept, as experiences, is a single “thing” having a specific structure. But that experienced structure is only weakly models as class and subclass relationships. Yes, I know that my dog “is-a” dog, but my experience of him has a far greater meaning.
From the point of view of my work, the (computer encoding of a) situational ontology should emerge from "something". It should allow "interpretation" and yet should also have a computable nature.
<Quote>
2.2.1: "Today's WWW is geared towards
presenting information to humans"
<end quote>
This statement is very different from the statement about "machines that understand each other". Human-centric information production and propagation is what is possible using machine encoding of ontological models. Computer data interoperability is related but has a difference nature, one that we can agree on and "walk forward", ie do something with the ontology.
The "something" advocated in the BCNGroup RoadMap is a subsetting of some, but not all, of the concepts in a domain and in a upper abstract ontology. SUMO, for me certainly qualifies as an upper abstract ontology - with concepts like: most general thing, process and location, for example. (Figure 1).
The report uses also the word/phrase "foundational ontology" and "universal ontology" . The Suggested Upper Merged Ontology (SUMO) is small, less than 50 concepts. (Sorry, I am still trying to see all of the concepts and count them.)
In developing the Roadmap, my group wanted to have an is-a property between elements of the upper abstract ontology and any element of the Mid-level and Domain ontologies. We then got the nice result that an ontology was a set of concepts which were sufficient to "cover" the description of a domain. The domain of all ontologies was then the upper abstract ontology. If concepts begin to appear in domain ontologies that did not map well to one or more concepts in the upper abstract ontology, then a modification was negotiated using collaborative software like SchemaLogic’s SchemaServer.
SUMO works as an upper abstract ontology, because the concepts in SUMO cover the "semantic space". The cover has the property that any reasonable concept can be discussed in the abstract using a subset of the concept elements in cover.
My work on this stopped when I could not get LMCO management to functionally provide to my team of ontologists a working LAN and the semantic extraction software which was to be the measurement device for constructing situational ontology about specific events.
Sorry to get into the details this way, but for our purposes we found SUMO to be sufficient as an abstract cover for things like Harmonized Tariff Schedule taxonomy (like the UDEF taxonomy) - as I have begun to illustrate at a series of communications I am posting at:
http://www.ontologystream.com/beads/nationalDebate/287.htm
The concept of a semantic cover has been developed in my unpublished work, and I am in a contract position to re-expose the notion of a topological cover , and its roots in Soviet era applied semiotics (Finn and Pospelov)... as well as the naturalness of the notion. Both the set of elements in a semantic topological cover and a (substructural) framework are enumerated to have the properties that a relative independence exists between pairwise enumerated elements and that the set of elements "cover" the space. In this case the "space" is not a logical or algebraic space - but a semantic space.
My feeling is that some of the objectives that John is referring to go away if there is a strong operational separation between the levels of ontology.
This strong operational separation between levels occurs IF the set of concepts in the upper abstract ontology (which SUMO is one) is a minimal and yet complete semantic cover. The domain and mid-level ontologies would then all be regarded as not at the same level of abstraction as the SUMO.
Mathematics is an upper abstract ontology and mechanical engineering is a mid-level and/or domain ontology. So it is easy to be clear. Any mechanical engineering concept has a specific "is modeled by" relationship to mathematics.
The situational, or scoped, ontology is a subset of abstract upper ontology, mid-level ontology and data instances; sufficient to providing and human the information sufficient to good interpretation (of data about a possible biological threat.)
Note:
I have to express the opinion that John Sowa's objections:
<quote>
> While there is an analysis cost in selecting an
upper ontology
> as theoretical framework, there is a greater
cost in not doing so,
> especially when one is dealing with relatively
abstract concepts.
What cost?
How has that been determined?
Which abstract concepts?
Abstract concepts are the most difficult to define
and the most
likely to vary from one ontology to the next. Wouldn't it be better
to avoid making any commitment to the details of any
version?
<end quote>
does not take into account the notion of a semantic cover.
SUMO provides a semantic cover at the upper abstract level... so why not simply work with this and move on?
Paul Prueitt