Database Reference
In-Depth Information
malformed Linked Data. While tool developers will have to contend with Linked
Data that does not conform to the Semantic Web best practices covered in this
chapter, we must hope that the semantic value of the content however will be
accurate. In the next chapter, we look at the issue of Semantic Spam and what hap-
pens when a data publisher has willfully produced incorrect data.
7.12 S U M M A R Y
This chapter has provided an introduction to how to publish Linked Data. We
explained Berners-Lee's Linked Data Principles, in particular the importance
of making URIs dereferenceable (i.e., it should be possible to look up the RDF
description of the resource identified by the URI on the Web). We have taken
the reader through the process of creating Linked Data: first, deciding what the
Linked Data is about; second, examining the current GI data and identifying what
should be included, left out, or modified; third, designing the RDFS ontology for
the data, deciding on the purpose, scope, and competency questions for the dataset
and reusing vocabularies if possible; fourth, minting the URIs for the data; and
finally, generating data by running additional GIS queries if necessary. The RDF
data can be published as a static RDF/XML file, as RDFa embedded in an HTML
Web page, or using one of the many software tools to create a Linked Data view
on a relational or RDF database. The publisher's own vocabularies should be made
dereferenceable, and each URI should be provided with an
rdfs:label
to enable
visualization tools to present a nice “human-readable” version of the resource.
Metadata should be attached to the dataset, preferably using VoID, to refer to other
access methods like RDF dumps and SPARQL endpoints, to specify what the data
is about (e.g., by including the purpose, scope, and competency questions), and
to include provenance and licensing information. Finally, to be really considered
proper Linked Data, the publisher must link its data to other Linked Data on the
Web for it to be discovered. This is the problem of data integration, and we address
this in the next chapter.
NOTES
1. While in American English the # is known as a pound sign, the Linked Data community
have adopted the British English word for it—the hash.
2. Other conventions are also used, and at the time of writing, none was universally
accepted, so you need to be aware of the particular convention used by the publisher of
the dataset you are interested in.
3. The
S
in RDFS stands for schema. The use of the word
schema
rather than ontology is
due to RDFS's origin as a metadata language that pre-dates Linked Data. In this topic,
we have therefore emphasized that the product is an ontology by referring to it as an
“RDFS ontology.” More normally, you are as likely to hear people talking about “the
RDFS” or “RDFS schema,” meaning the ontology written using RDFS, whereas ontolo-
gies produced using OWL are commonly referred to as just “ontologies.”
