Database Reference
In-Depth Information
Vo cabu la r y
20
takes a triple-centric perspective. It is a set of terms for describing
changes to triples, where the “Changeset” is the difference between the old dataset
and the updated one. There are also ontology-level change description mechanisms,
such as the OWL 2 Change Ontology (Palma et al., 2009), which allows the ontology
publisher to describe how one version of an ontology differs from its predecessor.
The Protégé ontology authoring tool offers the Change and Annotation Ontology
(Noy et al., 2006), which allows the data provider to specify changes to the ontology,
as well as version control information such as author of the change, timestamp, and
other annotations.
The DSNotify framework implemented by Popitsch and Haslhofer (2010) monitors
Linked Data sources and notifies applications consuming that data when the dataset
has changed. Links may be either structurally broken (when the object resource is no
longer retrievable) or semantically broken (where the link is semantically incorrect,
for example, two resources are linked with
owl:sameAs
when they are not in fact
describing the same thing). DSNotify assists with the detection and repair of broken
links by detecting structurally broken links and notifying the source data owner of
changes to target datasets and can also be configured as a service that automatically
forwards requests for moved target data resources to their new locations.
sparqlPUSH
21
is an interface that sits on top of a SPARQL endpoint, which
allows the specification of a number of SPARQL queries into the dataset denoting
which resources should be monitored and then uses the PubSubHubbub (Fitzpatrick,
Slatkin, and Atkins, 2010) real-time Web protocol to broadcast updates to the RDF
data store.
8.9
EVALUATING LINK QUALITY AND AVOIDING SEMANTIC SPAM
8.9.1
a W
ord
on
a
ccuracy
It is an open question regarding how accurate or specific we can realistically expect
Linked Data links to be. Many RDFS ontologies are quickly constructed, with the
express aim of not trying to express the finer details of semantics, and
owl:sameAs
is frequently used to indicate any form of similarity or relatedness, without the
author necessarily subscribing to or even fully understanding the implications of the
Description Logics equivalence relation. Some sources are more trustworthy than
others, and efforts under way into the expression of provenance, as discussed in
Section 7.7, will no doubt help link architects to verify sources and improve link
accuracy. However, it must be recognized that we are operating in a Web environ-
ment, and just as you do not assume that everything you read on the Web is true, you
perhaps should not expect to believe the results returned from a Linked Data query
that crosses multiple datasets. Future research might well look into how to increase
the accuracy of query results; if, for example, the same information was repeated in
multiple locations, this would lend credence to its veracity. In the meantime, systems
must be designed that do not require or expect perfect data and treat information
encoded in triples as claims rather than facts. In short, our advice is to be strict with
your output and tolerant with your input.
