ontology-summit
[Top] [All Lists]

Re: [ontology-summit] Defining "ontology"

To: "Ontology Summit 2007 Forum" <ontology-summit@xxxxxxxxxxxxxxxx>, "Deborah L. McGuinness" <dlm@xxxxxxxxxxxxxxxx>
From: "Obrst, Leo J." <lobrst@xxxxxxxxx>
Date: Wed, 24 Jan 2007 16:36:00 -0500
Message-id: <9F771CF826DE9A42B548A08D90EDEA800190B5B9@xxxxxxxxxxxxxxxxx>
Not quite, AJ: see some comments below.
 
I hope this helps a bit.
Leo
 
_____________________________________________
Dr. Leo Obrst       The MITRE Corporation, Information Semantics
lobrst@xxxxxxxxx    Center for Innovative Computing & Informatics
Voice: 703-983-6770 7515 Colshire Drive, M/S H305
Fax: 703-983-1379   McLean, VA 22102-7508, USA
 
 


From: ontology-summit-bounces@xxxxxxxxxxxxxxxx [mailto:ontology-summit-bounces@xxxxxxxxxxxxxxxx] On Behalf Of AJ Chen
Sent: Tuesday, January 23, 2007 2:40 PM
To: Ontology Summit 2007 Forum; Deborah L. McGuinness
Subject: Re: [ontology-summit] Defining "ontology"

I read Deborah's paper, Leo's presentation and other materials on ontology spectrum that were mentioned on this list. I found it's a nice way to view how knowledge representation has been evolved. If I understand correctly, the ontology spectrum view implies the term "ontology" is used almost as the replacement for "knowledge representation". 
 
[LEO:]  Not quite. The focus on the Ontology Spectrum/Semantic Continuum is more on the expressiveness of the "semantic model", from low to high. On the Semantic Continuum diagram, vocabularies and models are presented (on the ontology spectrum, if you extend the lower left below Taxonomy you would be in the realm of vocabularies; I chose to exclude those because my focus was on models -- although again, one could in fact introduce mathematical ordering/structure (it's implicit in the diagrams, but one can consider the ascending models as increasingly more structured). Though "knowledge representation" is closely related, I would prefer to keep KR as part of AI, where it began. 

The problem is that these diagrams are pedagogical and thus simplify things a bit, focusing as they do on the question of "what is an ontology", rather than on the pure expressivity or complexity of the model or logic (and most of the "semantic models" in the ontology are not based on logics). For my purpose (and I think for the purposes of Deborah McGuinness, Mike Uschold, Mike Gruninger, Chris Welty, and Fritz Lehmann), the overriding concern was to differentiate a range of less to more expressive models, some of which can be called "ontologies". For my purpose, it was to sort out what wasn't an ontology (logical theory), what could be an ontology (conceptual mode), and what wasn't an ontology (thesaurus, taxonomy). And thereby teach folks about these notions. However, these are simple diagrams and as such eliminate some complexity they really shouldn't, if one wants to really clarify these notions in some detail.

Some detail one could differentiate:

1) Orderings or structure. This is mathematical order/structure. Typically assuming set theory or category theory, one can define increasing "levels" of order. Using set theory, it is probably something like: sets, ordered pairs, partially ordered sets, ... lattices, ... Where the "order" relation/s is typically defined by mathematical properties suchas reflexivity, symmetry, transitivity, etc. With category theory, you get simple structural notions of categories (nodes) and morphisms (arrows), which you can generalize to functors, natural transformations, cartesian closed categories, topoi, etc.

2) Formal. Whether there is a language or logic (and a logic is a formal language) associated with the model, i.e., again rather simplistically, is there a formal syntax and a formal semantics for the modeling language? Sometimes we call this a "formal" model. In general, taxonomies and thesauri, or rather the languages they are expressed in, are not formalized, though one could develop a formal language for each. This formalization must be based on (1). Necessarily related to this is the notion of "machine-interpretability" and thus "automated reasoning/inference." If there is no logic behind the model, then the machine cannot "semantically interpret" that model, and thus cannot "automatically reason/infer."

3)  Philosophically ontological. Whether the model (ontology) is rooted in reality. And then what is the philosophical perspective of that rooting in reality? I.e., realist, idealist, nominalist, conceptualist, etc. I would also say that formal semantics comes into play here because we use symbols of a language and represent whatever those real-world referents are that those symbols refer to, leading to an intermediate realm of "concepts" or representations for those referents.

4) Use cases. What kinds of applications do these models address? For example, if you want to place documents into gross topical buckets (doc X is about China), for navigation or search that is a bit more structured than free-text (i.e., Google) search, then use a topic/term taxonomy. If you need great precision (i.e., you want to specify, generate, or find a specific service or application), you need a conceptual model or logical theory. If you want to have automated reasoning support for the latter, you need a logical theory.

So, the notion of semantic model or ontology brings together: mathematics, logic, theoretical computer science, AI,  formal language theory, formal semantics, formal ontology,  etc. As you can see, it is difficult to correlate all these sciences (and their related engineering or applied disciplines) in a simple single-dimensional spectrum focused just on "ontology". But I am optimistic and think that, though difficult, it is possible, especially if the purpose is pedagogical, and the result will help disparate communities that address similar "semantic" notions converge on common vocabulary and meaning for characterizing what they are about and how they are systematically distinguishable from others.

In addition to the broad view from "ontology spectrum", I'm interested in looking deeply in the more strictly defined ontology space, where an ontology represents a specific aspect of the real world using class objects that can be expressed in clearly defined languages (like RDF, OWL). By focusing on this strict ontology space, I hope we may be able to come up with some sort of framework that will help people develop more reusable ontologies (in RDF/OWL or newer languages). A classification scheme may be one component of this framework.

There are probably lots of work done in this direction already that I don't know about. Please provide pointers to any relevant work if you know any. Coming from object-oriented software development background, my sense is that we should be able to learn a lot from reusable software design. When an ontology is coded in a language, it's conceptually possible to map the ontology to a layer of ontologies designed with different granularity. And these different layers of ontologies should be about to talk to each other in an integrated application. 
[LEO:]  I think we come full-circle on this, since most of OO came out of AI and theoretical CS, i.e., from Smalltalk and similar languages, and the early pre-Common LISP notions of OO (Flavors, LOOPs), i.e., declarative representations of real-world objects, ways of structuring those objects, message-passing and eventually methods for communicating between objects and doing work, data abstraction and hiding, method/type coercing, inheritance, etc.

Again, using examples of real ontologies to further understand the problem of reusability is neccessary. In my previous post, I tried to use my own examples to illustrate the problem.  Hope more people will bring their examples into the discussion as well.

AJ
--
AJ Chen, PhD
http://www.web2express.org
"Open Data on Semantic Web"

Quoting "Deborah L. McGuinness" <dlm@xxxxxxxxxxxxxxxx>:

> hi - i am catching up on email so sorry to come in mid stream. i have
> not gone back through the whole thread but figured i would chime in now
> while the discussion is on a spectrum.
> i have gotten a LOT of leverage out of the notion of an ontology spectrum.
>
> In 1999 a number of us were on an ontology panel at aaai. as homework
> for the panel, we had a pre-meeting about what each of us were willing
> to call ontologies. It might be useful to explain my perspective when i
> participated (which actually still reflects my perspective now on the
> topic).
> this was just after i had done a fairly comprehensive consulting project
> looking at "naturally occurring" things that might be considered
> ontologies as a LARGE crawl to obtain starting points for a very
> comprehensive ontology.
> In that effort i looked in excruciating detail at a number of mostly
> taxonomies like yahoo shopping, lycos, amazon, as well as a number of
> subject specific taxonomies and things i consider light weight
> ontologies - class hierarchies where the classes have a small number of
> properties, sometimes with value restrictions. i was working in the
> context of a startup with a very broad user base so somewhat by
> necessity, i needed to consider what the general public might consider
> an ontology and how a broad set of people might use it.
> i also was in the midst of a large ontology-driven project - hpkb -
> where my team had to answer questions using 80 kbs as input. the
> knowledge bases were all essentially kif statements (or they had been
> translated into kif) and they were generated by people at least
> reasonably trained in kr and the kbs had a lot of structure. so i had
> been doing A LOT of ontology and knowledge base merging at that time in
> my life - an it was driven somewhat by two fairly different desires and
> needs for ontologies - 1 heavy weight theorem proving to generate
> answers and 2 very light weight ontology-enhanced search. (i still find
> myself driven by these 2 needs on a very regular basis).
>
> the ai panel members (welty, gruninger, uschold, lehman, and myself)
> came up with an ontology spectrum that i have grown to like quite a bit
> - but of course that is because it reflects a lot of my personal
> experiences with structured declarative knowledge representations from
> extremely lightly structured things to very principled, detailed
> structured artifacts. the belief i came to then and the one i still
> stand by now is that we are more likely to have structured knowledge get
> used in applications if we work with the notion of a spectrum and help
> people move along it as their needs and education permits.
>
> I gave a talk in early 2000 about the pull i was experiencing for
> ontologies at a dagstuhl meeting and used the spectrum as an organizer.
> I wrote a paper describing my view of each of the points on that
> spectrum along with some examples of my experiences in each of those
> points. i wrote the paper in 2000 with a small update in 2001 but the
> actual published version of it came out in a book from that even MUCh
> later - actually 2003.
> an online "preprint" version is up at:
> http://www.ksl.stanford.edu/people/dlm/papers/ontologies-come-of-age-mit-press-(with-citation).htm
>
> one reason i bring up that paper is that i still find that a lot of
> people tell me they get value out of that paper i think for 2 reasons -
> 1. the simple spectrum (yes - i think one dimension of expressiveness is
> really too limiting... but it is convenient pedagogically.)
> 2 the examples of each point on the spectrum.
>
> i would be happy to co-author a next generation of something like that
> paper.... and in fact, i have been asked for such a paper on a regular
> basis with more examples and more current references.
>
> deborah
>
> Obrst, Leo J. wrote:
>> Charles,
>> I agree with you. A number of us through the years have come up with
>> similar ontology continuums or spectrums. I prefer my Ontology
>> Spectrum*, but that's natural, I guess. It was developed over time to
>> act as an educational aid. I found that many folks understood notions
>> such as taxonomies, database schemas, UML models, but they didn't know
>> how these related to the new kid on the block, ontologies. Was a
>> thesaurus an ontology? No. Was a UML model: no, not yet. And term vs.
>> concept (placeholder for real world referent) is a crucial
>> distinction. The former is a word/phrase (string, utterance) that
>> indexes the latter, which is a representation of the meaning of that
>> term (at least approximately). The important point is that these
>> concepts/placeholders are meant to stand in for real world referents,
>> since ontology is about the things of the world. I also attach a newer
>> slide that tries to show those distinctions, along with their typical
>> use cases: OntologySpectrumApplication-Obrst06.jpg.
>> Thanks,
>> Leo
>> *If you look at the current Wikipedia article on the subject, it's not
>> completely accurate: http://en.wikipedia.org/wiki/Semantic_spectrum. I
>> independently developed the Ontology Spectrum in Fall, 1999, and it
>> really represents one dimension, though it is depicted diagonally (for
>> increased space) as though it were two dimensional: the one dimension
>> is in terms of expressivity of the model. Also the 4 way stations of
>> taxonomy, thesaurus, conceptual model, and logical theory are semantic
>> models; that is why I don't include glossaries, term lists, etc.,
>> directly -- they are not models but are human language lists and
>> definitions. Mike Uschold, Mike Gruninger, and Chris Welty and I have
>> talked about this topic of the co-invention of the semantic/ontology
>> spectrum for quite some time. Personally, I prefer my Ontology
>> Spectrum because I overlay onto the specific models additional
>> information, such as the kind of parent-child relation, related
>> database and modeling languages, and logic information. But all of
>> these ontology spectrum/semantic continuums are sound: they represent
>> the best distillations of solid generalizations especially good for
>> educational purposes.You are probably referring to the presentations I
>> gave at Ontolog last Jan 19/26 2006: "*What is an ontology? - A
>> Briefing on the Range of Semantic Models*",
>> http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2006_01_12.
>> _____________________________________________
>> Dr. Leo Obrst The MITRE Corporation, Information Semantics
>> lobrst@xxxxxxxxx Center for Innovative Computing & Informatics
>> Voice: 703-983-6770 7515 Colshire Drive, M/S H305
>> Fax: 703-983-1379 McLean, VA 22102-7508, USA
>>
>>     ------------------------------------------------------------------------
>>     *From:* ontology-summit-bounces@xxxxxxxxxxxxxxxx
>>     [mailto:ontology-summit-bounces@xxxxxxxxxxxxxxxx] *On Behalf Of
>>     *Charles D Turnitsa
>>     *Sent:* Monday, January 22, 2007 1:39 PM
>>     *To:* Ontology Summit 2007 Forum
>>     *Subject:* Re: [ontology-summit] Defining "ontology"
>>
>>     One of the big schisms in types of ontology that I see is a
>>     difference in an ontological representation (model) that is
>>     intended to organize knowledge at the level of terms, and a model
>>     that is intended to organize knowledge at the level of meaning.
>>
>>     If you look at the Ontology Spectrum that was presented to the
>>     Ontolog group last year by Dr. Leo Obrst, you see a progression of
>>     ontology representation techniques, from controlled vocabularies
>>     and simple data models, up through thesauri, taxonomy techniques,
>>     up to axiomatized systems and logic based models (and beyond). One
>>     of the big shifts I have seen is the difference in emphasis of
>>     lower level models (thesauri and controlled vocabularies, for
>>     instance) on terms, and the attempt of upper level models (axiom
>>     based systems, logic models) on definitions. For different
>>     communities, differently focused applications, both appear equally
>>     useful, but they are very different.
>>
>>     From all of this, possibly an axis of differentation for
>>     ontologies can exist to show the focus of what the ontology is
>>     defining, and the depth of it's intended use.
>>
>>     Chuck
>>
>>     Charles Turnitsa
>>     Project Scientist
>>     Virginia Modeling, Analysis & Simulation Center
>>     Old Dominion University Research Foundation
>>     7000 College Drive
>>     Suffolk, Virginia 23435
>>     (757) 638-6315 (voice)
>>     (757) 686-6214 (fax)
>>     cturnits@xxxxxxx <mailto:cturnits@xxxxxxx>
>>
>>     -----ontology-summit-bounces@xxxxxxxxxxxxxxxx wrote: -----
>>
>>         To: Ontology Summit 2007 Forum <ontology-summit@xxxxxxxxxxxxxxxx>
>>         From: Patrick Durusau <patrick@xxxxxxxxxxx>
>>         Sent by: ontology-summit-bounces@xxxxxxxxxxxxxxxx
>>         Date: 19/01/2007 08:53AM
>>         Subject: [ontology-summit] Defining "ontology"
>>
>>         Greetings,
>>
>>         I am concerned with the suggestions that it is possible to
>>         create a
>>         continuum along which to organize what are known as
>>         "ontologies" in one
>>         or more circles.
>>
>>         At least unless we are willing to concede that the creation of
>>         such a
>>         continuum is itself an imposition of assumptions from an
>>         undisclosed
>>         ontology.
>>
>>         I am sure there are those who would say that folksonomies are
>>         "missing"
>>         features that are present in "formal" ontologies. Perhaps, but
>>         folksonomies predate "formal" ontologies by several millenia
>>         and have
>>         proven robust enough for many purposes. If the goal is to
>>         represent the
>>         opinions of the many rather than the few, perhaps it is "formal"
>>         ontologies that "missing" features.
>>
>>         I am not taking a position one way or the other. But, I do
>>         think it is
>>         important to realize that any attempt to construct a continuum
>>         is with
>>         an unstated choice of a winner before the the continuum is
>>         populated.
>>
>>         Hope everyone is looking forward to a great weekend!
>>
>>         Patrick
>>
>>         --
>>         Patrick Durusau
>>         Patrick@xxxxxxxxxxx
>>         Chair, V1 - Text Processing: Office and Publishing Systems
>>         Interface
>>         Co-Editor, ISO 13250, Topic Maps -- Reference Model
>>         Member, Text Encoding Initiative Board of Directors, 2003-2005
>>
>>         Topic Maps: Human, not artificial, intelligence at work!
>>
>>
>>
>>         _________________________________________________________________
>>         Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/
>>         Subscribe/Config:
>>         http://ontolog.cim3.net/mailman/listinfo/ontology-summit/
>>         Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx
>>         Community Files:
>>         http://ontolog.cim3.net/file/work/OntologySummit2007/
>>         Community Wiki:
>>         http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2007
>>         Community Portal: http://ontolog.cim3.net/
>>
>>
>>
>> ------------------------------------------------------------------------
>>
>> ------------------------------------------------------------------------
>>
>>
>> _________________________________________________________________
>> Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/
>> Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/
>> Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx
>> Community Files: http://ontolog.cim3.net/file/work/OntologySummit2007/
>> Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2007
>> Community Portal: http://ontolog.cim3.net/
>>
>
>
> _________________________________________________________________
> Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/
> Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/
> Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx
> Community Files: http://ontolog.cim3.net/file/work/OntologySummit2007/
> Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2007
> Community Portal: http://ontolog.cim3.net/
>




_________________________________________________________________
Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/ 
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/  
Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx
Community Files: http://ontolog.cim3.net/file/work/OntologySummit2007/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2007
Community Portal: http://ontolog.cim3.net/    (01)
<Prev in Thread] Current Thread [Next in Thread>