Re: Using Linking Open Data datasets

2008/5/30 Richard Cyganiak <richard@cyganiak.de>:
> Peter,
>
> On 29 May 2008, at 22:10, Peter Ansell wrote:
>>
>> It is a very good idea IMO to start putting sparql endpoints into RDF
>> descriptions of datasets. If you also indicate a typical triple
>> pattern that can be found at that endpoint/graph tuple you instantly
>> enable basic distributed sparql based on SPARQL GRAPH, where you
>> derive the information from a markup and not the expertise of the
>> querier.
>>
>> I think providing descriptions at the level of each SPARQL named graph
>> available at different endpoints is appropriate if people use GRAPHS
>> for macroscopic categorisation and not use a different graph for each
>> URI they loaded for instance. I know there is no standard governing
>> the use of SPARQL GRAPH's but it would be nice to have at least large
>> blobs available where you are most likely to find data matching a
>> triple pattern, whether it be DARQ's predicate idea (bad IMO) or some
>> combination of predicates with subjects and objects.
>
> A partial solution to this are Semantic Sitemaps [1]. They provide for
> discovery of SPARQL endpoints to RDF datasets, but don't have the detailed
> descriptions. There is a way of defining a dataset URI though, which could
> be used to provide additional meta-information in RDF.
>
> Richard
>
> [1] http://sw.deri.org/2007/07/sitemapextension/

That looks very usable to me. Has anyone used it for linked data? How
do you discover these sitemaps as a linked data user, as opposed to
sitemaps which are traditionally submitted to search engines for
searching. In either case, it would be nice to have an RDF description
submitted as part of a sitemap to a semantic search engine so it might
be good to standardise that mechanism based around these ideas.

Also, there is a reference in that document to N-Quad format, what is
that exactly? [2] is a bit sparse on examples so it is hard to
understand what is meant by the syntax.

Also, is the slicing declaration attempting to make up for a deficit
in the SPARQL protocol w.r.t. DESCRIBE? Why not utilise SELECT if you
had an idea of what pieces of information you desire, although I guess
the server is in the best position to recommend information to you
with DESCRIBE queries. I think slicing mechanisms should be defined
outside of that context, although the lack of progress with CBD [3] is
a little worrying with respect to that bit.

[2] http://sw.deri.org/2008/02/nx/
[3] http://www.w3.org/Submission/CBD/

Cheers,

Peter

Received on Thursday, 29 May 2008 23:02:41 UTC