W3C home > Mailing lists > Public > public-rdf-wg@w3.org > April 2012

Re: regrets and a new spin on contexts

From: Pat Hayes <phayes@ihmc.us>
Date: Mon, 30 Apr 2012 00:40:11 -0500
Cc: public-rdf-wg@w3.org
Message-Id: <06315CF6-0921-4BDF-B956-C5C6D861035D@ihmc.us>
To: Antoine Zimmermann <antoine.zimmermann@emse.fr>

On Apr 29, 2012, at 5:47 AM, Antoine Zimmermann wrote:

> Le 29/04/2012 05:08, Pat Hayes a écrit :
>> Sorry, I have been offline for a few days. I will tidy up and expand
>> the wiki page ASAP, probably Monday.   In the meantime:
>> 
>> On Apr 25, 2012, at 9:02 AM, Antoine Zimmermann wrote:
>> 
>>> The semantics does not say to what syntax it applies. There are
>>> examples in TriG given, but does it mean that the semantics apply
>>> to RDF Datasets?
>> 
>> To the graphs in the dataset. However, I was using Sandro's
>> convention for distinguishing between genuine graph names and mere
>> dataset labels, and that needs the dataset syntax.
> 
> So it does not define a formal semantics of datasets. So the question is still open: do we want to define a formal semantics for datasets or not?

I have been assuming that datasets have essentially the very minimal semantics suggested by Sandro, in which the default graph is asserted and the remainder of the dataset asserts that a labeling relationship holds between the entities denoted by the 'name' IRIs and the graphs they are associated with in the dataset, and that this relationship amounts to a genuine naming relation when appropriate RDF is contained in the default graph. (Whether or not the named graphs are also asserted, and whether or not the named graphs are complete or partial, are both irrelevant to this proposal.) 

The spirit of the proposal is to provide a way to interpret RDF graphs which provides for the graphs in a dataset to have the interpretation, when considered simply as graphs, that you proposed for a 'named' graph inside a dataset. So the idea would be to say that a dataset is simply a collection of graphs (which is what they appear to be, to the naive viewer) rather than an entirely new syntactic construction with a sharply different semantics. 

> 
>>> This proposal does not make explicit what is the satisfaction
>>> relation (the ⊨ symbol, \models in LaTeX
>> 
>> Thanks, but I am allergic to latex.
>> 
>>> ) that relates interpretations to the theories in the logic that
>>> they satisfy.
>> 
>> I was trying to avoid too much mathematical flimflammery, but I will
>> write it out more exactly.
> 
> It is the most important part. It provides the technical details that allow one to implement it. In order to be able to compare it to the other existing proposal, you need to get the definitions straight. People of this working group, all being highly educated I suppose, understand maths.

I think I should refrain from comment at this point. 

> 
>>> 
>>> It says: "the rest of the semantics (for triples, graphs, blank
>>> nodes, etc.) are exactly as in the 2004 semantic specifications"
>>> 
>>> but this is not enough. As an example, this is not sufficient to
>>> determine what happen to bnodes (especially bnodes that appear in
>>> different "named" graph).
>> 
>> The semantics applies to graphs, so the blank nodes really are
>> exactly as in 2004. We could change this, of course, but that is
>> orthogonal to this issue.  It uses the Trig machinery only to attach
>> names to graphs, since these names now can play a stronger semantic
>> role.
> 
> Names are syntax. You can't say that we need names to define the semantics and at the same time say that names are not in the syntax.

Of course. (? I am not sure what your point is here.) All I meant was that there are two orthogonal issues: how to attach names to (named) graphs - which I am here assuming is done using the dataset syntax - and how to interpret graphs in contexts, which I am suggesting be done by a different syntactic convention than the one you proposed, and incorporated into the RDF graph machinery rather than into the SPARQL dataset machinery. So datasets are simply a way of 'baptising' graphs, and the graphs themselves can 'import' or 'inherit' their contexts. However, the actual context semantic machinery is exactly like the one you were proposing, and indeed was essentially copied from your proposal.

>>> There's also the term rdf:Graph used in one example but nothing is
>>> said about how it is interpreted in the model theory (but ok, it's
>>> just an example).
>>> 
>>> Yet anyway, with what's written I am able to extrapolate and what I
>>> see is essentially the same as my proposal in [1], on top of which
>>> the idea of "rdf:inherits" has been built.
>> 
>> Yes, it is partly based on your idea. It amounts to a generalization
>> of it along the lines of the Cyc-Guha-McCarthy context logic. (I
>> stated this explicitly in the earlier draft which uses the "context"
>> terminology, see
>> http://www.w3.org/2011/rdf-wg/wiki/RDFwithContexts#Two_current_ideas
>> .)  It differs in that it the rdf:inherits provides a way to
>> distinguish between graphs being merely named and those being
>> asserted in a "context". I think it is important to keep this as an
>> option, as this allows both your kind of intended uses (with a small
>> syntactic overhead, admittedly, which I think can be minimized) and
>> those in which people strongly do not wish to contextually-isolate
>> their IRIs.
> >
>> BTW, I intended this to be seen as as extension of your semantic
>> idea, incorporating it into the RDF model itself. It changes the
>> graph semantics so as to make a dataset (modified slightly using
>> rdf:inherits and graph names) have your semantics as an option, that
>> option being controlled by RDF content in the graphs.
> 
> Ok let's see, but my proposal was to make the semantics of datasets distinct from the semantics of RDF and apply minimal changes to the semantics of RDF.

And in that way, certainly, this differs from your proposal. It amounts to a more substantial change to RDF itself. On the other hand, I feel that it provides a more coherent overall picture, and it makes the context machinery (or whatever we want to call it) more widely and usefully available, since "contexts" can be referred to explicitly and linked to in the linked-data sense, rather then being restricted to use inside a single dataset. Once this idea (and the similarity of context inheritance to owl:imports) occurred to me, I became enthusiastic about how generally useful it might be for the whole semantic web, so this idea has now somewhat grown beyond the narrow question of giving a semantics for datasets. 

It may be that your context genie has gotten more outside the bottle than you had intended it should :-)

Pat



> 
> 
> AZ
> 
>>> 
>>> 
>>> 
>>> Also a side remark wrt the form: I would be happier if the
>>> semantics use "I ⊨ G" instead of "I(G) = true", as according to my
>>> experience, there is an overwhelming majority of authors in the
>>> knowledge representation field (especially semwebers) using the
>>> former and not the latter. It would also allow us to avoid
>>> introducing the unorthodox notion of "occurrence". But this may
>>> just be a question of taste.
>> 
>> I think this can be phrased to avoid the "occurrence" language, which
>> I agree is awkward. Im working on it.
>> 
>> Pat
>> 
>> 
>> 
>>> 
>>> 
>>> 
>>> [1] Dataset semantics.
>>> http://www.w3.org/2011/rdf-wg/wiki/TF-Graphs/RDF-Datasets-Proposal#Semantics
>>> 
>>> 
>>> 
>>> 
>>> 
> Le 23/04/2012 09:45, Pat Hayes a écrit :
>>>> First, regrets for next Wednesday, I will be driving through
>>>> Texas. Second, I have written up essentially the same proposal in
>>>> a slightly different terminology which might (?) be more
>>>> palatable, anyway it is there for inspection at
>>>> http://www.w3.org/2011/rdf-wg/wiki/AnotherSpin
>>>> 
>>>> Pat
>>>> 
>>>> ------------------------------------------------------------
>>>> IHMC (850)434 8903 or (650)494 3973 40 South Alcaniz St. (850)202
>>>> 4416   office Pensacola                            (850)202 4440
>>>> fax FL 32502                              (850)291 0667 mobile
>>>> phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>> 
>>> -- Antoine Zimmermann ISCOD / LSTI - Institut Henri Fayol École
>>> Nationale Supérieure des Mines de Saint-Étienne 158 cours Fauriel
>>> 42023 Saint-Étienne Cedex 2 France Tél:+33(0)4 77 42 83 36
>>> Fax:+33(0)4 77 42 66 66 http://zimmer.aprilfoolsreview.com/
>>> 
>>> 
>> 
>> ------------------------------------------------------------ IHMC
>> (850)434 8903 or (650)494 3973 40 South Alcaniz St.
>> (850)202 4416   office Pensacola                            (850)202
>> 4440   fax FL 32502                              (850)291 0667
>> mobile phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes
>> 
>> 
>> 
>> 
>> 
>> 
> 
> 
> -- 
> Antoine Zimmermann
> ISCOD / LSTI - Institut Henri Fayol
> École Nationale Supérieure des Mines de Saint-Étienne
> 158 cours Fauriel
> 42023 Saint-Étienne Cedex 2
> France
> Tél:+33(0)4 77 42 83 36
> Fax:+33(0)4 77 42 66 66
> http://zimmer.aprilfoolsreview.com/
> 
> 

------------------------------------------------------------
IHMC                                     (850)434 8903 or (650)494 3973   
40 South Alcaniz St.           (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes
Received on Monday, 30 April 2012 05:40:48 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 22:02:04 UTC