[Top] [All Lists]

Re: [ontolog-forum] Named Graphs, Statements, and Triples

To: ontolog-forum@xxxxxxxxxxxxxxxx
From: John F Sowa <sowa@xxxxxxxxxxx>
Date: Sat, 14 Jun 2014 17:51:13 -0400
Message-id: <539CC3D1.8010909@xxxxxxxxxxx>
Kingsley and Frank,    (01)

This new name for the thread is a better description of the issues.
But see the note by Mark Linehan from the old thread (copy below).
After that is a shortened version of my earlier note.    (02)

> 1. Named Graph IRIs denote distinct RDF data sources which may be
> local or external to a DBMS -- basically, they are Document/Database
> identifiers    (03)

> In short, my interpretation is that a Named Graph is nothing more
> than a "collection" that has an identifier and a set of traits.    (04)

At that level of detail, I agree with both Kingsley and Frank.  But
I believe it's important to distinguish three independent notions:    (05)

  1. Some notation that groups or encloses the data.  Peirce used an
     oval to enclose any graph or subgraph of interest.  IKL (which I
     mentioned and Mark endorsed) uses an operator named 'that' plus
     parentheses to enclose the sentence or Boolean combination of
     sentences of interest.    (06)

  2. Some notation for referring to that group.  In various logics,
     it can be a global name (such as an IRI) or a local reference
     (variable, pointer, or position in the syntax).    (07)

  3. The semantics for using that grouping.  In both existential graphs
     and IKL, the oval or the 'that' operator acts like quotation marks
     that suspend further evaluation of the content.  That enables any
     operator that processes the group to specify its own version of
     semantics for interpreting (or not interpreting) the data in it.    (08)

> 2. Reified Statements (i..e., Statements denoted by an IRI) are
> instances of rdf:Statment meaning they are the domainOf rdf:subject,
> rdf:predicate, rdf:object properties    (09)

The idea of reification is an artifact of the assumption that only
a physical object named by a noun can be a first-class "thing"
(or 'res' in Latin).    (010)

Quine was one of the philosophers who promoted that very limited
style of ontology.  I like to cite a talk that Alonzo Church
delivered at Harvard -- with the intention of annoying Quine:    (011)

    The Ontological Status of Women and Abstract Entities
    http://www.jfsowa.com/ontology/church.htm    (012)

Peirce had a much simpler approach:  All content in any ontology
is represented by relations with zero or more arguments.  There is
only one quantifier:  a line that says "Something exists."  A name
is a relation that is true of something by convention.    (013)

For higher-order logic, Peirce allowed references to relations.
But he also allowed the option of defining any relation as a graph
that contains some combination of lines and other relations.  If you
replace a relation name with its defining graph, you can refer to
(link to) any line in it.    (014)

This is an extremely simple, general, and flexible method that
avoids mysterious terminology -- like reification.    (015)

> 3. Triples being the content (in the form of subject, predicate,
> object patterns) of RDF documents which are denoted by IRIs
> (typically  HTTP URLs when external to the DBMS).    (016)

For generality, I would replace the word 'triple' with arbitrary
sentence in whatever logic is used.  If it's a graph logic, that
would mean any graph or subgraph that can be considered to have
a truth value.    (017)

The W3C has accepted JSON as an option for representing RDF.
That's a step toward greater generality, since JSON is not
limited to triples.    (018)

But you need a foundation that is as clean, simple, and general
as possible.  And that foundation must be capable of supporting
*everything*.  That includes the Semantic Web, RDF in all its
variations, all the logics the work with RDF, *and* all the legacy
systems and their semantics (formal or informal).    (019)

I'm glad to see Mark's endorsement of IKL for specifying SBVR.
That's another logic that has not yet been brought into the
SW family, but it's within the scope of Tim B-L's vision.    (020)

For an intro to existential graphs that illustrates the points
above, see http://www.jfsowa.com/talks/egintro.pdf    (021)

The first dozen slides in egintro.pdf are enough for an overview.
The later slides get into more detail.  For even more detail, see
the references in slide 40, especially egtut.pdf and eg2cg.pdf.    (022)

For IKL, see http://www.jfsowa.com/ikl/    (023)

John    (024)

-------- Original Message --------
Subject: Re: [ontolog-forum] Requesting Opinions on the Benefits
of Predicates as Nodes
Date: Fri, 13 Jun 2014
From: Mark H Linehan    (025)

Yes, I agree with what John says.    (026)

SBVR (the OMG "Semantics of Business Vocabulary and Rules" standard)
comes at semantics, ontologies, and rules from the perspective of
business users and business usage. It is defined using "SBVR Structured
English", which is a form of Controlled Natural Language.  An important
(but under-defined from a logician's point-of-view) concept in SBVR is
"state of affairs", which is the class of instances of propositions.    (027)

So one can define "employment" as the state of affairs that a company
employs a person, and then one can define attributes of the
"employment" concept -- i.e. attributes of the class of instances
of any proposition that uses "company employs person".    (028)

When Ed, I, and others defined a "Date-Time Vocabulary" for SBVR,
one of the key requirements was to relate time to concepts such as
"employment".  We used "state of affairs" for that.  We also mapped
the key aspects of the Date-Time Vocabulary to Common Logic and IKL,
using the "that" operator to express the relationship between
propositions such as "company x employs person y" and attributes
such as start date.    (029)

Net: the IKL "that" operator is key to addressing a requirement
that arises in modeling real business vocabularies and rules.    (030)

Mark    (031)

-----Original Message-----
From: John F. Sowa    (032)

Jack and Mark,    (033)

> If I might add, I will assert that predicates as first class citizens
> is perhaps the only difference between an RDF graph and a topic map.    (034)

> Consider the triple {CompanyX employs Person1} using a predicate
> called "employs".  The triple implicitly describes what an English
> speaker would conceptualize as an "employment", with attributes such
> as "start date", "end date", a location, etc.    (035)

Yes.  This is an important issue, and it's obscured by an unfortunate
tendency of some logicians to call verbs predicates.    (036)

Peirce observed that when you say "Mary gave a child a book" there are
four entities involved:  Mary, the child, the book, and the act of
giving.  He allowed any or all of the four entities to have its own
existential quantifier (or "line of identity").    (037)

For conceptual graphs, anything you might want to refer to or link
to is represented by a concept node.  Anything you don't intend to
refer to may be represented by a relation node.  But you always have
an option of analyzing any relation as a graph with more nodes that
can be referred to or linked to.    (038)

For the KR ontology (http://www.jfsowa.com/ontology/ ), the default
mapping is to assign a concept node to every verb.  That node has
an implicit quantifier, which represents the action or the state.    (039)

The idea of using the letters SVO to refer to the three parts of
an RDF triple is more confusing than helpful.  I don't recommend it.    (040)

John    (041)

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    (042)

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