- From: Harry Halpin <hhalpin@w3.org>
- Date: Tue, 14 Dec 2010 18:11:32 -0000 (GMT)
- To: public-rdb2rdf-wg@w3.org
I think this may solve some of the Linked Data aspects. Michael points
about quite a bit more, but this is just about reusing URIs.
In particular, it seems the simplest solution would be to, given a string
from a database, to retrieve a full URI from linked data or some other
source. All that is needed is a standard REST-ful interface for passing a
string and getting back a URI.
This could be something as simple as this:
http://www.example.com/r2rml?string=Cambridge+MA
with results being a URI, i.e.
http://dbpedia.org/resource/Cambridge_MA
Then the URI of the REST API call could then be a parameter set in R2RML
(or even direct mapping). The string should be sent encoded as a URI query
parameter as normal.
If one wanted more than one response per string, then we could recommend
using a vocabulary like OpenSearch [1] or a RDF equivalent thereof, or
just a list of URIs delimited by carriage return or tab and another
variable in the call for getting the top, say, 12, which would then look
like this:
http://www.example.com/r2rml?string=Cambridge+MA&results=10
With different result formats, you could say:
http://www.example.com/r2rml?string=Cambridge+MA&results=10&format=txt
http://www.example.com/r2rml?string=Cambridge+MA&results=10&format=rdf
http://www.example.com/r2rml?string=Cambridge+MA&results=10&format=xml
This approach could then be plugged into OKKAM, Sindice, some vocabulary
host, etc. and so is neutral.
cheers,
harry
[1] http://www.opensearch.org/Specifications/OpenSearch/1.1
Received on Tuesday, 14 December 2010 18:11:33 UTC