Re: First order logic and SPARQL

On Tue, Sep 7, 2010 at 1:19 PM, Axel Polleres <axel.polleres@deri.org> wrote:
> Indeed, we have implemented only the other direction, SPARQL to Datalog, as a prototype plugin for the
> dlvhex system cf. http://dlvhex-semweb.svn.sourceforge.net/viewvc/dlvhex-semweb/
>
> Why would you need the translation of Datalog into SPARQL? That is something IMO just needed for the theoretical expressivity result,

This direction is useful for rule-based query re-writing.  For
example, If you want to answer the following SPARQL:

SELECT * { ?person a :Man }

And you have the following datalog rules:

Man(?X) :- hasSex(?X Male), age(?X ?AGE), math:greaterThan(?AGE 18)

You can leverage a logic programming theorem prover / reasoner (such
as Prolog) to unify the query against the rule, and generate the
following SPARQL query to answer the question (via translating the
body into a SPARQL query):

SELECT ?person
{
  ?person hasSex Male;
       age ?AGE
  FILTER(?AGE > 18)
}

To do this, you will need to be able to do the translation in a way
that preserves the semantics, of course.

>you dan't want to translate your datalog program into SPARQL and evaluate it with a sparql engine, that is likely not going to be very efficient...

If you do this in a naive fashion, yes.  But presumably, you could
take advantage of optimizations in the theorem prover (such as
re-using bindings provided in the query, for instance) and/or other
query planning statistics such as the size of the predicates to
reorder the body before converting it into a SPARQL query.  Also, this
approach doesn't rule out the evaluation of the SPARQL in an
efficient, mature, and even remote, engine.

-- Chime

Received on Tuesday, 7 September 2010 17:44:05 UTC