Database Reference
In-Depth Information
7.11.2 c
orrect
UrI D
ereferencInG
Regarding correct URI dereferencing, this is the question of whether the Linked
Data server performs a 303 redirect on the URI correctly and eventually returns
the description of the RDF resource. A number of tools can check this for you.
The Vapour Linked Data Validator
65
provides a report about the HTTP requests
and responses that took place as the URI was looked up. The URI Debugger
66
is a
similar tool for debugging Linked Data sites and URI dereferencing. Hurl
67
makes
basic HTTP requests, so you can see what is happening as your URI is dereferenced.
RDF:Alerts
68
is another Web tool to validate RDF/XML; it dereferences the URI and
checks the data is valid against its RDFS ontology. If all else fails, there is also the
command-line-based tool cURL,
69
which transfers data in URL syntax to and from
a server; this basic functionality allows you to check the fine detail of the HTTP
communication to ensure the 303 redirects are operating correctly.
7.11.3 c
ontent
: S
eMantIc
c
orrectneSS
To test whether the Linked Data set is fully interconnected, it is worth using a
Linked Data browser such as one of those discussed in Section 7.10. Check whether
every URI in the dataset can be reached by browsing along the links. A fully inter-
connected Linked Data set will assist the Linked Data crawlers in indexing all of the
dataset. Load the data into a tool that can perform semantic inferencing (a browser
like Tabulator can do basic RDFS inferencing) and check that the data conforms to
the domain and range restrictions, and no URI is inferred to be an instance of the
wrong class. Hogan et al. (2010) reported on the types of errors frequently found
in Linked Data, which included properties found in the object/value position of a
triple; misuse of
owl:ObjectProperty
(see Chapter 9), when a datatype property
should have been used instead; and the hijacking of properties defined elsewhere for
the authors' own purposes. The worst example of this last transgression was when
the standard property
rdf:type
was redefined.
7.11.4
c
ontent
: S
eMantIc
c
oMpleteneSS
For checking semantic completeness, convert the competency questions for your
Linked Data into SPARQL (see Chapter 8) and query them using a SPARQL endpoint
on your data to check that the data returns all the answers you expected. If it does
not, you may be missing data resources, but more likely, you will be missing some
links between resources. Check the individual triples in Turtle format as this is a
more concise, easy-to-read syntax.
7.11.5 e
xpect
the
W
orSt
The analysis of errors in Linked Data from Hogan et al. (2010) suggests that Linked
Data on the Web must be
expected
to be buggy, and just as browsers today are
forgiving of errors in HTML and users do not necessarily believe everything they
read on the Web, so will the Linked Data tools of the future have to be forgiving of
