Re: Blank nodes as predicates Re: Blank Nodes Re: Toward easier RDF: a proposal

> On 23 Nov 2018, at 12:40, Ross Horne <ross.horne@gmail.com> wrote:
> 
> Hi Henry,
> 
> On Fri, 23 Nov 2018 at 08:59, Henry Story <henry.story@bblfish.net> wrote:
> The interesting thing seems to be that when one does that, if one takes blank nodes to be
> existential quantifiers, as I think we all do, then blank nodes fall out automatically in
> predicate position too.
> 
> Surely the language that falls out of RDF by adding existential quantifiers in predicate position is a fragment of SPARQL ASK (before adding disjunction and equality).
> 
> E.g.: The following is a fine SPARQL Query (is :a related to :b and :c in the same way):
> ASK { :a ?x :b . 
>           :a ?x :c }

That is a question whereas a graph is a statement. Given the answer True to that question
you can complete your ASK query to generate the minimal graph - with a bnode in predicate
position - to make the statement that fulfills the query. That graph would be

   :a _:rel :b.  :a _:rel :c.

This is because truth is disquotational. "Snow is white" is true iff Snow is white, is a
well known statement on which deflationary theories of truth have been build.
https://plato.stanford.edu/entries/truth-deflationary/ In N3 one can write the rule

{ { ?a ?r ?b } a log:Truth } log:implies { ?a ?r ?b}

So your ASK sentence goes beyond the IT signature morphism mentioned in
   https://www.iep.utm.edu/insti-th/#SH3c
that I have adapted to ground rdf graphs because it is
  1) dialogic 
  2) it has a relation on graphs (the relation  ?g a log:Truth)

The signature morphism in IT that maps RDF ground graphs to RDF with existential
quantifiers is a very simple one Σ → Σ＋X . I am happy to take that part out to my thesis
and go over this in more detail to show how it works. Perhaps I can turn it into a peer
reviewed paper? (I am thinking that cutting my thesis up into shorter papers may be a
better strategy as it will be easier to find reviewers for parts of the thesis than for
the whole one).

> The closer RDF and SPARQL are, the better. Perhaps, if the development of RDF and SPARQL had followed more unifying language-design principals (e.g. institutions), design mismatches may have been reduced. However, now history has run its course, it's hard to see how institutions can achieve impact here. Please correct me if you believe otherwise.

From the first chapter of IEP article mentioned above

 "Institution theory was introduced by Joseph Goguen and Rod Burstall in the late
 seventies as a response to the explosion in the population of logical systems in use in
 formal specification theory and practice. Formal specification is a logic-based area of
 computer science that aims to support reliable system development through axiomatic
 formalisation of their structure and functionality. At the time (and now even more) there
 was a great diversity of specification formalisms, each of them supported by a particular
 underlying logical system. Hence the need for a uniform approach to specification theory
 capable to develop those part of the theory that are independent of the choice of a
 particular logical system, and thus are common to many specification formalisms. The key
 step was the definition of the concept of institution in (Goguen and Burstall, 1984)
 intended to capture formally the structural essence of logical systems beyond specific
 details. Since semantics plays the primary role in formal specification, institutions
 lean towards the semantics side of logic, known as model theory."

The idea of IT was to bring together a vast number of different logics in existence at the
time. RDF has been developed by many very good people, so I think it unlikely that a huge
problem has arisen here if it has been found possible to find a common structure to all these
other logics that were not even designed in the same instution. On the other hand it could 
be very helpful to make the links between these different pieces crystal clear.

> 
> There may be some hope. For example, Andy Seaborne's idea of a high level language for Linked Data, navigational in style [1], has not yet been take up by the W3C. Such a language might benefit from a more unifying semantics, in which standards encapsulated are precisely related, informing language-design decisions.
> 
> Kind regards,
> 
> Ross
> 
> [1] http://w3-org.9356.n7.nabble.com/SPARQL-friendly-alternative-to-rdf-Lists-td266182.html

Received on Friday, 23 November 2018 15:03:44 UTC