Re: [namedgraphs] [Semantic Web Client Library] T-Box and different sources

[Olaf H. <ha...@in...>]

Hello Cristina,

On Monday 20 September 2010 07:52:55 Cristina Sarasua wrote:
> Hi,
> 
> I would like to ask something about the Semantic Web Client Library.
> 
> I've been executing some SPARQL queries using the code specified in [1].
> 
> With the following query:
> 
> String queryString =
>             "PREFIX foaf: <http://xmlns.com/foaf/0.1/> " +
>             "SELECT ?i ?p WHERE {" +
>             "?i ?p <http://www.w3.org/People/Berners-Lee/card#i> . " +
>             "}";
> 
> I get resources from different sources.
>
> But, with this query:
> 
> String queryString =
>             "PREFIX foaf: <http://xmlns.com/foaf/0.1/> " +
>             "SELECT ?i ?p WHERE {" +
>             "?i ?p foaf:Person . " +
>             "}";
> 
> all the results belong to the same source - the document where foaf is
> implemented.

That's a result of the way the SWClLib executes queries. It uses a new 
approach for query execution which we call link traversal based query 
execution. The SWClLib Web site outlines the idea of this approach. Did
you read how this approach works?
You find a more detailed description of this approach in my paper at
ISWC 2009.

So, what the query engine does for your example query is the following: Given 
we start with an empty queried dataset (a.k.a. cache). SWClLib looks up the 
URI for foaf:Person, retrieves the RDF graph that defines the FOAF vocabulary, 
and adds this graph to the queried dataset. Now it tries to find RDF triples in 
the queried dataset that match the triple pattern in your query. That's only 
triples from the FOAF vocab graph (because that's the only one we have). The 
FOAF vocab graph does not contain RDF links that SWClLib can follow (where 
"following" here means looking up URIs to retrieve further RD graphs). Hence, 
SWClLib cannot discover more data. That's always a problem with this query 
approach when the query contains only URIs for vocabulary terms (or t-Box 
concepts as you call them) but no URIs for individual things (i.e. instances 
of classes defined in vocabularies).

Notice, however, that I added Sindice support to the SWClLib. With Sindice 
support enabled the SWClLib does not only look up URI (using the HTTP 
protocol) but it also queries the Sindice search engine for additional RDF 
graphs that contain RDF triples with the URI. This practice may give you more 
results for your second query. But, it's very expensive in terms of query 
execution time. Therefore, Sindice support is disabled by default. The Web 
site documents how to enable it.
 
> I wonder whether the Semantic Web Client Library does not retrieve
> resources from different sources when querying about T-Box, or whether
> this is just a coincidence.

It's not about t-Box or a-Box. To the SWClLib everything are just RDF graphs, 
SWClLib does not do any kind of reasoning.

Greetings,
Olaf

> Thanks in advance.
> 
> Kind regards,
> 
> Cristina
> 
> [1] http://www4.wiwiss.fu-berlin.de/bizer/ng4j/semwebclient/#using