W3C is pleased to receive the Concise Bounded Description from W3C Member Nokia (see submission form). The Concise Bounded Description Submission defines an algorithm for describing resources by selecting a portion of the RDF graph near the resource.
A service that receives a query about a resource may respond with any of three flavors of the Bounded Description described in the submission. The SPARQL Query Language has provisions for a DESCRIBE operation where the graph returned is at the discretion of the service. It is likely that some form of bounded description will be a useful way to answer many of these queries. Further, these bounded descriptions may be used to locate subgraphs for update or deletion.
While other forms of bounded description are discussed, this submission promotes CBDs which select statements coming from the node
being described, i.e. statements having the described node as the subject.
This subject traversal seems arbitrarily asymmetric; for instance, an owl:FunctionalProperty
can distinguish subject nodes just as an owl:InverseFunctionalProperty
can distinguish object nodes. This asymmetry prevents CBDs from being used
for general knowledge discovery. Interesting links are just as likely to have
the queried node as an object as a subject.
Currently, owl:FunctionalProperty
is rarely used to
unambiguously identify the object of the property. A brief survey of existing
RDF applications (FOAF, CC/PP, EARL, XML, Creative Commons, Annotea)
indicates that the subject traversal of CBDs is useful for indirectly
identifying objects, i.e., tying them to real-world entities. A CBD
of a foaf:Person
, for example, identifies the person by
revealing the person's name and email address. These applications are,
however, not predictive of future semantic web applications. The semantic web
is formed by the merging of many often tree-like applications. The
concept of which information it is natural to expect in a query may be
application-specific. It is possible that reliance on CBDs will bias
vocabulary developers who wish to make sure their data is included in CBDs. This
cost should be considered before they are adopted for general use.
It is easy to construct a graph where a bound description is impractically large. However, in given applications, CBDs may be useful. It is possible that, after evaluating use cases and considering various inclusions and exclusions from the CBD, their will be a practical way to consistently respond to generic resources queries. In an effort to control the size of the graph, the submission suggests assigning re-usable identifiers to what would otherwise be bNodes in the graph. However, the reader should be aware that raises issues related to httpRange-14.
While some query languages allow a requester to specify the exact arcs to be included in a return graph, some allow the server to define the return graph. SPARQL's DESCRIBE operation is mentioned above. Other protocols use similar forms of semantic discovery, for example TAP and Annotea. Using a form of bounded description would give clients more certainty of what the returned graph would include.
The RDF Data Access Working Group (DAWG) is approaching completion of SPARQL. Implementations of SPARQL will provide valuable implementation experience of DESCRIBE. This experience may lead to further standardization work in the area of semantic discovery.
Disclaimer: Placing a Submission on a Working Group/Interest Group agenda does not imply endorsement by either the W3C Staff or the participants of the RDF Data Access Working Group, nor does it guarantee that the Working Group will agree to take any specific action on a Submission.