SUO: The lattice of theories
Pierre-
Just a couple comments to clarify/answer your questions ...
-FF
At 11:10 2002-06-11 +0200, Pierre Grenon wrote:
>
> Franck,
>
> I'm not sure whether I understand the registry thing. Are you talking about
> developping a standard ontology? (I thought the mere idea was banished from
> this group ;) ...
I mentioned in a previous E-mail the idea: sometimes it takes a while for to determine the "right" problem to solve -- regarding SUO, my thinking has changed over the past couple years. So the problem has changed from "developing a standard upper ontology" to "developing a standard framework for ontology access/use", which means (in my mind): registration (the registry), expressing *certain* relations in a standard way, and having consistency across the descriptive aspects of each entry in the registry ("metadata for ontologies").
In another area of activity that I participate in, healthcare, there has been a conference series that has been going on for the past decade or so called "Towards and Electronic Patient Record" (TEPR). They don't call it the "Standard Electronic Patient Record", even though many conference presenters and the conference itself cared about the topic. I believe that had they called it "Standard Electronic Patient Record", they would have continally declared failure every year (how discouraging!) because, after a decade, they still don't have a standard patient record (it's not clear exactly what the scope of that standard might be -- e.g., just medical records, or maybe scheduling within hostpitals, too?).
So the administrative "brilliance" (or luck) of TEPR was to recognize the incremental nature of the work. With a title like "Towards ...", TEPR does accomplish its goal of getting closer and closer (most people involved in healthcare standardization recognize that it is a long-term, consensus-building process).
Back to your question: Am I talking about developing a standard ontology? No, not exactly. I'm talking about an incremental process ... there may be some useful/useless good/bad things that get registered. Let the user/reader choose. Maybe a standard ontology does arrive, but let's not make it a requirement (yes, my thinking has changed).
To use John Sowa'a library analogy: the nature of cataloging, shelving, searching, and selection in a library are pretty much the same across all books, but some books are better/worse than others, more/less useful than others, more/less popular than others, more/less truthful than others, etc.. A distinction between a library and a registry might be: the library, typically, has some minimal policy on what is acceptable or not (e.g., a musical library might not want books on sports) and there is very little "interaction" among books in the library (e.g., if book P states X and book Q states ~X, then this is typically not a worry among librarians).
A registry might have (via its registration authority process) a policy that has stronger requirements on what may be registered. Or maybe there are several kinds of registration possible:
- a consensus-building registration (strong registration authority process): there has been some consensus-built around these entries; the consensus-building process is referenced/documented
- a private-use registration (weak registration authority process): anyone can register; access to the private-use may be public access, limited access, no access, etc.
Certainly, some private-use registrations can be popular. And certain submitors for private-use registrations may have their own internal criteria for what they consider to be an acceptable registration.
While this approach doesn't guarantee a "standard", it does work "towards" more consensus.
As a side note: about a month ago, I mentioned my interest in the ISO 31-* series on units of measure. This summer, I plan on attempting to code some portions of that ISO standards work into KIF (I'm making preparations now). Will this become part of some "standard" ontology? I don't know and I'm not too worried. Given that it is an ISO standard, won't it be part of a "standard" ontology? No, not necessarily -- does SUO WG decide this kind of thing? I'm more concerned about representing/coding something where there is already strong international agreement. Will it be useful? I hope so.
> ... I guess my question should be: would you describe Type 1
> standards as ontological and Types 2 and 3 as meta-ontological?
In your terminology, Type 1 would be "ontological" and Type 3 would likely be "meta-ontological" (just cataloging information), but I'm not yet sure about Type 2.
> ... It might be
> usefull for instance to explain what you have in mind when you refer to the
> process of using concepts as foundations for developping further concepts.
> Do you mean that further concepts are defined or are framed in terms of
> those belonging to the intial registry? I'm curious about this framing.
> Would concept be introduced as primitive and used in some rules together
> with earlier concepts or would they be described in terms of the earlier
> concepts (there would be a mix of ontological and meta-ontological content
> in the registry). Note: I'm using the concept terminology here, and rather
> loosely I guess, I'm not really familiar with this and not very confortable
> either.
I hope your "concept terminology" is consistent with my "terminology terminology". :-)
I would expect that some entries in the registry are dependent on other entries in the registry. There are a variety of ways to characterize this kind of relationship. For example:
Illustration #1:
P: aslkj daslaskj aslk aslka dlkaj dalskdj
Q: aslkjsa sj sdk sj ask akd asdasd
R: ak ask ak askaskjda dal ad $P ajk $P dask d
S: aksj asjd asdjs $Q asjkd s $R akjdkasd
Illustration #2 ("<" = someone dependent on me, ">" = I'm dependent on someone)
P [<R]: aslkj daslaskj aslk aslka dlkaj dalskdj
Q [<S]: aslkjsa sj sdk sj ask akd asdasd
R [>P]: ak ask ak askaskjda dal ad ajk dask d
S [>Q >R]: aksj asjd asdjs asjkd s akjdkasd
In the above illustrations, when viewed from the perspective of the entries, the dependency relationships among *entries* are the same in #1 and #2, but within a different relationship perspective, such as context and use, it would reveal a different kind of relationship structure, e.g., in #1, R uses P twice in the context of "ak ask ak askaskjda dal ad $P ajk $P dask d" (whatever that means), while in #2, R may refer to P, but we don't know anything else about that reference.
I don't yet know what all the features of an entry might be.
More importantly, what is or what can be an entry? A concept, an axiom, a theory, an ontology, a relationship, a statement, "just KIF code", etc.? I'm using the term "entry" because it has a convenient lay meaning ... and I don't have to get tied up in "what an entry is". However, I would like to explore this topic via a discussion in SUO WG: what are the fundamental units that might be registered? What do we call them? For now, "entry" seems OK, with fewer arguments. Anyone with suggestions?
-FF
_______________________________________________________________________
Frank Farance, Farance Inc. T: +1 212 486 4700 F: +1 212 759 1605
mailto:frank@farance.com http://farance.com
Standards, products, services for the Global Information Infrastructure