Re: Vocab AND Ontology?

From: Bernard Vatant <bernard.vatant-at-mondeca.com>
Date: Mon, 24 Sep 2007 18:06:21 +0200


Hi Tony
>> documents/data about "White Dwarf Stars", a SKOS concept "White Dwarf
>> Stars" is what you need, and the position of this concept in a Concept
>> Scheme hierarchy, and its association with other concepts such as
>> "Chandrasekhar Limit".
>>
>
> I don't recall any concept linking proposed for vocabs apart from the
> narrower/broader relationships.

In the IAU thesaurus, like in most thesaurus, there is the notion of "Related Terms" (RT) otherwise called "associative relationship". It is non hierarchical. Basically it means : if you are interested in this, you should also see that.
> What would you term this relationship
Related Terms (RT) in thesaurus standards. In IAU thesaurus you have this :

CHANDRASEKHAR_LIMIT

        D
                "Chandrasekhar limit"
        RT
                GRAVITATIONAL_COLLAPSE
                MASS
                MASS_RADIUS_RELATION
                WHITE_DWARF_STARS



> and how would it be recorded in SKOS?
sv:Chandrasekhar_limit skos:related sv:Gravitational_Collapse sv:Chandrasekhar_limit skos:related sv:White_Dwarf_Stars
etc
> I assumed this sort of relationship was
> better modelled in an ontology.
>

Actually not at all. This is exactly the kind of relationship with fuzzy semantics (even more fuzzy than broader/narrower) extremely useful for human search and retrieval, but hard to encode in formal logic. The meaning of such a relationship depending on what you want to achieve with it ...
>> - If you want to classify objects, find them using logical inference,
>>
>
> I'm not sure we do want to classify individual objects so much as classify
> datasets, queries, papers, people even?
What is the purpose of Ed's Star Ontology in this case, if not for classifying individual stars? I mean here representing individual stars as instance of a specific subclass of star. This is the functional objective of such an ontology. Otherwise I don't understand what a "star" class and its subclasses is good at. Certainly not at classifying papers, data, observations about stars, people and labs working on them, whatever - which would need other ontologies - an ontology of documents, an ontology of data, of observations etc ...

I
>> But you don't find "Chandrasekhar Limit" at all in the star
>> ontology, although it's a very pertinent concept in stellar
>>
>
> I know I'm exposing my lack of knowledge about ontologies, but that's what
> these lists are for - to learn something new from others. Why can we not add
> "Chandrasekhar Limit" as a concept in an ontology?
You certainly can, in (too) many ways. I'm curious to know how Ed would fit this concept in, and what logical articulation with White Star class he would propose.
> Does every concept have to be a class of physical things? Can things not have a relationship with
> this concept of either 'beyond' or 'within'?
>

What "things" do you speak about here? Individual stars? or Star classes?. If you want to express formally something as : "A white star has a mass beyond the Chandrasekhar limit", it's a necessary condition you put on the "white star" class. And what kind of "thing" is "Chandrasekhar Limit"? A concept? The value of a property? And if yes, which property? It's a numerical value, right? If you look closely at OWL, you will find it extremely difficult to express that kind of relation, because it involves a specific numerical value, with numerical constraints (mass greater than, etc ...) ... for which OWL, despite its great expressivity, has no real satisfying representation (First Order Logic doesn't like continuous quantities).

So actually, trying to express constraints like "The mass of a white dwarf is within Chandrasekhar's limit" which is AFAIK a way to declare in natural language the relationship between the two concepts (correct me if I am wrong - this is certainly approximative and even wrong for experts, but what they will propose instead will be even more difficult to capture I'm afraid), there are two ways. The SKOS / Thesaurus fuzzy way, saying those two concepts are closely related, and when one speaks about theformer, he's likely to say something about the latter, or the OWL hard way, and well, good luck.
> BTW - some of the things I've read before:
> 'A formal ontology is a controlled vocabulary expressed in an ontology
> representation language.'
> (http://www.metamodel.com/article.php?story=20030115211223271)
>
> 'ontologies are specified in the form of definitions of representational
> vocabulary'
> (http://informationr.net/ir/6-2/paper94.html)
>
> 'map the relations between differing vocabularies, and pinpoint the location
> of properties within a larger ontological framework of interconnected
> knowledge, interconnected via the relationships established within an
> ontology'
> (http://xml.coverpages.org/ni2001-02-28-d.html)
>

I will go and read those in more details bandwidth permitting, but none seems contradictory with what I say above. Vocabularies are indeed generally the basis for ontology construction, concepts don't pop up out of thin air, they are most of the time embedded in natural language, and the ontology miner work is to try to explicit whenever possible their implicit semantics. But there is a "whenever possible". It's not because "white dwarf" and "Chandrasekar's limit" concepts occur in the same vocabulary context that expliciting the semantics of this co-occurrence is straightforward, or even possible in a given representation framework. And having more expressivity does not necessarily help. Or it helps to discover that you've better stick to fuzzy relationships ... :-)
> I'd be interested in your take on these documents (and many others beside)
> and why you think, either they've got it wrong or they're talking about
> something different than we are in this thread. I'm not trying to dispute
> your expert opinion but showing how I came to the conclusion that a
> vocabulary might be contained within an ontology.
>

I have no disagreement with this conclusion. But devil is in the details. Some terms/concepts are straightforward to represent formally as classes (Star, Planet), and some as properties (mass, radius) and some as individuals (Sun, Moon, Messier 31). And some are less so. Chandrasekhar's limit is one of those.

Bernard

-- 

*Bernard Vatant
*Knowledge Engineering
----------------------------------------------------
*Mondeca**
*3, cité Nollez 75018 Paris France
Web:    www.mondeca.com <http://www.mondeca.com>
----------------------------------------------------
Tel:       +33 (0) 871 488 459
Mail:     bernard.vatant-at-mondeca.com <mailto:bernard.vatant-at-mondeca.com>
Blog:    Leçons de Choses <http://mondeca.wordpress.com/>
Received on 2007-09-24Z18:06:26