ontolog-forum
[Top] [All Lists]

Re: [ontolog-forum] using SKOS for controlled values for controlled voca

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
Cc: Graeme Hirst <gh@xxxxxxxxxxxxxx>
From: Ed Barkmeyer <edbark@xxxxxxxx>
Date: Tue, 12 Oct 2010 12:19:08 -0400
Message-id: <4CB48A7C.3050509@xxxxxxxx>
Whilst not claiming to the education that you all clearly possess in 
this area, I would like to support John's position, if in a curious 
way.  I do not believe that many textbooks are written in natural 
language, and I am not convinced that many formal documents -- treaties, 
contracts, political positions, etc. -- are written in 'natural language'.    (01)

It seems to be a human trait, or at least a European one, to create very 
formal languages that resemble natural language, but have purely 
technical vocabularies, and carefully chosen turns of phrase that are 
themselves technical in meaning.  Yes, these writings follow the natural 
language grammar, but tools that support the truly 'natural' language, 
like WordNet, are nearly useless in determining the semantics of the 
utterances.    (02)

In short, many of the corpora that we are now trying to interpret 
automatically are written in controlled languages, with some natural 
language properties.  And it seems to me that the primary use of SKOS is 
to capture the fixed terminology of such languages.  But SKOS lacks the 
ability to deal with juxtapositions and turns of phrase, which are 
common parts of the means of conveying the semantic load ("context", if 
you must) in these controlled languages.    (03)

Conversely, it has been my experience that almost all ontologies are 
subjective, which is to say, they represent the objective world as seen 
through a particular lens.  In that way, they have all of the properties 
of a semantic corpus that we assume for a natural language text, and 
document with some kind of "metadata" -- author, date, etc.  But whereas 
a natural language text can betray its viewpoint internally, by choice 
of terms and positional linkages, ontologies explicitly assume that a 
term means its formal definition and nothing more.  So, as Leo says, 
there is a continuum here, and it is probably better to think of it as a 
multidimensional surface than a 2D curve.  Every point in that continuum 
has some value in capturing and conveying knowledge, but the values may 
be of different kinds.     (04)

The problem of classification of these points (a la Linnaeus) is to 
carefully define the specific observable traits that are necessary and 
sufficient conditions for membership in a category.  The best we can say 
for ontologies is that they are written in controlled languages that 
have a formal mathematical semantics, full stop.  That definition is 
very clear.  Whether it says anything of real value about the knowledge 
conveyed is 'quite another thing entirely'.  It says nothing whatever 
about their 'semantics' in the linguistic sense, or their 'viewpoint', 
etc.  Ontologies are yet another form of conveying knowledge, that has 
both the advantages and the limitations of technical support.    (05)

-Ed    (06)




John F. Sowa wrote:
> Patrick, Graeme, Leo, Simon,
>
> There are two totally different ways of using logic + ontology:
>
>   1. Artificial mathematical systems:  Developing a system of formal
>      logic and using it to define some kind of artificial system, over
>      which we have total control.  An example would be some system of
>      pure mathematics for which we have no applications in mind.
>      Another example would be a some data structures inside a digital
>      computer.
>
>   2. NL semantics:  Describing some subject that people talk about
>      in some natural language.  This includes every branch of science,
>      engineering, medicine, business, politics, and the arts.
>
> For anything in #2, the entities described by the ontology must
> be related directly or indirectly to whatever people who use the
> terminology refer to.  You can't claim that one notation (say SKOS)
> deals with terminology and another one (say OWL) deals with ontology.
> If they are using the same character strings for the same subject
> matter, there is no difference.
>
> PD:
>   
>> Graeme has another observation about language that may be of interest:
>>     
>
> GH:
>   
>> This seems to arise from a combination of overenthusiasm for Western 
>scientific method
>> and a misunderstanding of the nature of language that borders on fear. In 
>this view, language
>> is a messy and highly imperfect medium that is not to be trusted, but rather 
>must either be
>> sidestepped entirely or be beaten into submission by means of logic and 
>formalism.
>>     
>
> LO:
>   
>> This seems like an anti-science and anti-logic quotation, I hope out of 
>context.
>> It seems it is hard to be a computer scientist or engineer if you don't 
>believe
>> in science and logic.
>>     
>
> I think that Leo doesn't get the point that Graeme is trying to make.
> My interpretation of Graeme's comment is that certain logicians, such
> as Frege, Russell, Carnap, and Montague, were hopelessly misguided
> about the nature of language and its relationship to logic.
>
> I completely agree with that interpretation.  My three favorite
> philosophers are Peirce, Whitehead, and Wittgenstein.  As logicians,
> they were just as brilliant as the four above.  But they also had
> a much deeper and more realistic understanding of the nature of
> language and its relationship to the world and to the people who
> use it to talk about the world.
>
> In my paper "The Role of Logic and Ontology in Language and Reasoning",
> I make the point that it is disastrously misleading to claim that the
> semantics of NLs is *based on* logic.  Instead, it is more appropriate
> to say that every version of logic and ontology that anybody has ever
> proposed has been invented as an *abstraction from* NLs.
>
> LO:
>   
>>  I don't think the answer is to throw our hands up and say
>> "Oh, language is ineffable, inscrutable!"
>>     
>
> I completely agree.  Peirce, Whitehead, and Wittgenstein would never
> say that.  See my paper with quotations from them.
>
> SS:
>   
>> Controlled Vocabularies are Controlled, in the same way that Controlled
>> Natural Languages are Controlled...
>>
>> However, the constraints that are placed on the Terms permitted/the subset
>> of syntax recognized require a certain degree of training before new terms
>> can be added, or new sentences written.
>>     
>
> I have been advocating CNLs for years. You can use CNL terms in
> an abstract way when you're defining an artificial mathematical
> system (point #1 above).  But when you're using logic plus ontology
> for #2 (anything in the real world), the controlled vocabulary is
> nothing more nor less than the labels for your ontology.  And it is
> also a subset of the same terms that are commonly used in the field.
> Some of the unused senses for each term might be disallowed in a CNL,
> but people routinely do that with any NL.
>
> SS:
>   
>> The important thing to remember is the KOS Concepts are essentially
>> intentional-with-a-t.
>>     
>
> You're making a distinction between prescriptive (intentional) vs.
> descriptive (presumably more objective).  So what?  You can use NLs,
> CNLs, or ontologies in either way.
>
> SS:
>   
>> Controlled Vocabularies that allow Hierarchical relations must
>> follow other constraints: the most important one is that the
>> relationship must always be true. That means that if Term A has
>> Broader Term B, everything that is about A must necessarily
>> in some fashion also be about B. [Term == Concept in KOS speak].
>>     
>
> That's fine.  But note the word 'about'.  That word is not
> spelled the same as the logicians' word 'denote'.  That
> indicates that they don't have the same meaning.
>
> For example, the term 'paw' denotes a part of an animal,
> not a type of animal.  But if you talk about a cat's paw,
> you are talking *about* a cat.
>
> The relation 'subtype', which is usually used in ontology,
> is defined in terms of denotation.  The relation 'narrower',
> which is commonly used in library science, is defined in
> terms of aboutness.  Denotation and aboutness are not the
> same.  Therefore, 'subtype of' and 'narrower than' are
> not the same.
>
> The fact that SKOS uses one term and OWL uses another means nothing.
> You could introduce 'aboutness' as a primitive into OWL and use it
> to define narrower-than.
>
> LO:
>   
>> Of course there is a continuum working here, as is usual, but there
>> can be discrete points in that continuum, which act as clarifying
>> spikes or way-stations...
>>     
>
> I certainly agree.  But the level of precision that is possible with
> any subject depends on the application.  SKOS is more often used
> with applications for which people make less sharp distinctions.
> If you use OWL for the same subject, you can't make it any more
> precise than you can with SKOS.
>
> LO:
>   
>> there are typically clear differences between controlled vocabularies
>> and thesauri, on the one hand, and ontologies, on the other.
>>     
>
> No.  Look at medicine, for example, there is a one-to-one mapping
> from the terminologies used by the physicians to the ontologies.
> Patients, however, don't know the more precise terms.  So they
> use vaguer words, and they don't observe the constraints.
>
> LO:
>   
>> "Narrower than" is not the same as "Subclass of"
>>     
>
> Of course not.  They are spelled with different character strings.
> That indicates that they are different.  Note my comment to Simon.
>
> So far, you have not given a single reason why SKOS can't be
> considered anything but a subset of OWL with a different choice
> of metalevel terms.
>
> John
>  
> _________________________________________________________________
> Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
> Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
> Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
> Shared Files: http://ontolog.cim3.net/file/
> Community Wiki: http://ontolog.cim3.net/wiki/ 
> To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
> To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx
>  
>       (07)

-- 
Edward J. Barkmeyer                        Email: edbark@xxxxxxxx
National Institute of Standards & Technology
Manufacturing Systems Integration Division
100 Bureau Drive, Stop 8263                Tel: +1 301-975-3528
Gaithersburg, MD 20899-8263                FAX: +1 301-975-4694    (08)

"The opinions expressed above do not reflect consensus of NIST, 
 and have not been reviewed by any Government authority."    (09)


_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (010)

<Prev in Thread] Current Thread [Next in Thread>