Re: Really minimal dataset semantics

Le 21/09/2012 15:46, Sandro Hawke a écrit :
> On 09/20/2012 12:00 PM, Antoine Zimmermann wrote:
>>> Huh? If the meaning of a named graph is tied up with relating names to
>>> graphs, then the semantics certainly has a role there.
>> Sorry, maybe I misunderstood what you were saying, but then I don't
>> understand your point.
>> What I'm saying is that, if you find a dataset somewhere in the wild,
>> or if you have a dataset in memory, you can get the graph associated
>> with a graph IRI by simply parsing the dataset representation.
>> Semantics does not come into play in that case.
> I don't see how that can work, exactly.

Dataset d = DatasetFactory.createDatasetFromFile(new File("foo.trig"));
Graph g1 = d.getGraph("");

That's it.

> You also said, stating the basic idea behind entailment:
>> The goal, the way I see it, is to define a semantics that satisfies
>> the requirements that we all agree on, such that if a conclusion is
>> drawn from a dataset by applying the semantics, no one would object to
>> that conclusion.
> I like that idea. Each module takes in some RDF (graphs or datasets),
> does good stuff to it (never violating the semantics), and then passes
> more RDF onward. As long as what came in was true, and the semantics
> were properly designed, what goes out will be true.
> But if what came in was a harvester dataset, where the named graphs
> contain exactly the triples parsed from the content obtained via Web
> dereferencing, then even with your most minimal semantics, the dataset
> coming out may well be full of falsehoods.

So, you find a dataset in the wild which happen to be using a vocabulary 
that indicates it's the result of a Web crawl. Having knowledge about 
this vocabulary, you know that the content is such that the named graphs 
contains, /verbatim/, what they found by looking up the graph IRIs. So 
your application simply assumes that it is indeed a Web crawl and 
consider the graphs in the <n,g> pairs to have been "quoted".

That does not involve semantics. If you are suspicious whether the 
crawler is behaving properly (that it is not changing the content before 
serialising it) you still check the source IRI, but even in that case, 
it does not prove anything. No semantics can ever tell you with 
certainty that the thing has indeed been quoted without modification.

If all you need is exactly *what's written* in the dataset or part of 
it, just read it. If what you need is *the knowledge* conveyed by a part 
of the dataset, then use the semantics. And the semantics does not do 
anything, it's just helping you *know* certain conclusions, these 
conclusions do not affect the data as they are written, unless you *want 
to modify* the data.

> (In other words, can you tell me again how ds-entailment works in a
> world where we sometimes want quoting semantics?)
> -- Sandro

Antoine Zimmermann
ISCOD / LSTI - Institut Henri Fayol
École Nationale Supérieure des Mines de Saint-Étienne
158 cours Fauriel
42023 Saint-Étienne Cedex 2
Tél:+33(0)4 77 42 66 03
Fax:+33(0)4 77 42 66 66

Received on Friday, 21 September 2012 15:28:54 UTC