[Top] [All Lists]

[ontology-summit] Offline note. Re: First Model Bench Challenge

To: ontology-summit@xxxxxxxxxxxxxxxx
Cc: bradley.shoebottom@xxxxxxxxxxxxx
From: John F Sowa <sowa@xxxxxxxxxxx>
Date: Sat, 19 May 2012 14:28:47 -0400
Message-id: <4FB7E65F.5040508@xxxxxxxxxxx>
Jack, Bradley, Cory, Amanda, Leo, Doug, et al.    (01)

I was working on the slides for my tutorial at the Semantic Tech Conf
in San Francisco while this thread was going on.  Since it is related
to my talk, I'd like to refer to the slides.  But I don't want to
send the URL to a public forum until after I present them on June 4.    (02)

However, I put the slides on my web site in a temporary directory
that is protected from Google by robots.txt:    (03)

     Knowledge Design Patterns    (04)

The basic issues are related to the huge amount of conflicting
metalanguage, as Jack noted:    (05)

Jack R
> We may have quite different labels for the same concepts...
> Am I understanding your intent?    (06)

Most of the conflicts are created by an overabundance of metalevel
jargon.  That's why I emphasize first-order logic.  It only requires
simple English words: and, or, not, if-then, some, and every.    (07)

If you use FOL as your metalanguage, you can get rid of talk about
classes, attributes, properties, and features.  Other terms that
create more confusion than enlightenment are 'universals' and
'particulars'.  You can talk about all the issues with just one
metalevel term:  *relation* .    (08)

In the kdptut.pdf file, see slides 39 to 67.  They show how simple
English words can be mapped to different notations with a minimal
amount of metalanguage.    (09)

Bradley S
> The Legacy also belongs to the object property or type "Compact"
> class of vehicles. This makes it a Sedan as compared to a SUV.    (010)

I don't disagree about the point, but you can get rid of the metalevel
jargon if you use FOL as your metalanguage.  In any dialect of CL,
you can use the word 'relation' to replace 'property', 'class', and
'type'.  You can even apply relations to relations.  The only other
useful metalevel term is 'function'.  See slides 66 and 67.    (011)

I admit there is a need to talk about collections of various kinds.
For that purpose, I suggest the word 'set'.  Sets are very widely
taught and used, and they create far less confusion than classes.    (012)

I also admit that the word 'type' is useful.  I suggest the
definition "a monadic relation used to determine membership in
a set."  (You can even get rid of Greek words like 'monadic' by
calling them one-place, two-place, or N-place relations.)    (013)

> I have used "type" to mean the logical concept (same as yours) and
> "class" to be the OO concept which also embodies an instance factory.    (014)

That's another reason for getting rid of the word 'class' when talking
about ontology.  The words 'relation' and 'set' are used in common
speech and in many different professions.    (015)

If you define a type by a one-place relation, you can avoid conflicts
with common usage and with most technical uses.  Aristotle's word
'category' is another option.  It has a long tradition in philosophy,
and TV game shows have popularized that usage.    (016)

> Having federation as a purpose gives us that guideline. If, say,
> we were providing semantics for demographic data, we might need
> one or more of the more specialized definitions.    (017)

Purpose is essential for anything related to life.  Social systems
and demographic data are certainly related to life.  For purpose,
see slides 70 to 78.  For systems, see slides 98 to 99.    (018)

Note to Leo:
Your comment about the need to avoid confusing semantics and ontology
is important.  I agree, but in an earlier comment, I said that the
word 'semantics' has already acquired many conflicting uses. I discuss
some of the issues in Section 6 (slides 84 to 101) of kdptut.pdf.    (019)

Note to Doug F:
Section 7 (slides 102 to 119) discusses my current views about
the relationship between NLs and controlled NLs.    (020)

I used to define a CNL as a formal language that uses a subset
of the vocabulary and syntax of some NL.  But I don't think it's
useful to make a sharp distinction.  I believe we can have tools
that support a continuum of gradually tightening constraints:
totally unconstrained tweets, somewhat better email, carefully
edited documents, CNLs written for human readers, and CNLs that
are just as formal as any computer language.    (021)

John    (022)

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/OntologySummit2012/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2012  
Community Portal: http://ontolog.cim3.net/wiki/     (023)
<Prev in Thread] Current Thread [Next in Thread>