- From: Enrico Franconi <franconi@inf.unibz.it>
- Date: Wed, 7 Sep 2005 03:47:21 +0200
- To: Bijan Parsia <bparsia@isr.umd.edu>
- Cc: RDF Data Access Working Group <public-rdf-dawg@w3.org>

On 7 Sep 2005, at 03:13, Bijan Parsia wrote: > Looking again at the charter: > """"1.8 Derived Graphs > > The working group must recognize that RDF graphs are often > constructed by aggregation from multiple sources and through > logical inference, and that sometimes the graphs are never > materialized. Such graphs may be arbitrarily large or infinite.""" > > Another way to look at this is that for some forms of "logical > inference" (the kind which generates a model), it's not just the > case that the graphs are never materialize, or arbitrarily large, > or infinite, but there might be *multiple* such graphs generated > from the same "base". > > So lets say you forget about deductive closure and focus on what > I'm going to call "ABox completion", where each ABox is an RDF > Graph generated by completion rules. Then you can sparql query by > subgraph *that*. Now, a hit has to be defined by the binding being > found in *ALL* completions. > > Each ABox corresponds to a model. Ergo, we are back to entailment > without much fuss. Yeah, I like it. This corresponds directly to the definition of what in the DB world and in the DL world is called "certain answer": whatever is true in any possible model. My example was explained in this way, by showing that an answer should hold in all models. Node that formally speaking checkiing that something is true in all model is exactly equivalent to check that it is entailed. Now we have three possibilities: 1) replacing "subgraph of" with "entailed by"; 2) explaining that the subgraphing is done on the deductive closure of the original information (clumsy); 3) explaining that the subgraphing on all the models of the original graph (requires some work to find the proper wording). I still prefer (1) for its simplicity and generality. --e.

Received on Wednesday, 7 September 2005 01:47:31 UTC