The semi-structured nature of Semantic Web data and its use of multiple vocabularies make it challenging to describe the structure of some RDF dataset. When trying to understand the content behind some SPARQL endpoint, users usually need to write a number of SPARQL queries to \mine" the types contained in the dataset and their relations. We argue that this is a tedious, error-prone and time consuming task to be carried out manually. SPARQture, the tool described in this paper, automates this process and renders the structure of a particular RDF data available behind some SPARQL endpoint using a number of visually rich components. Moreover, SPARQture supports user interaction to further explore the data. We describe SPARQture and discuss its t to the IESD Challenge. The reader is encouraged to view the demo at: http://sparqture.appspot.com
The Semantic Web community succeeded in making large amounts of structured
data available on the Web in the last decade. This data, represented using the
RDF data model, spans various domains from social networks, to bioinformatics
to geography. Many of these datasets are made available via SPARQL endpoints.
SPARQL [
SPARQL endpoints are commonly accessible over the Web. They can be queried using any tool that supports the HTTP protocol or via an HTML form, typically with a large input text. The user is then left with the daunitng task of discovering the content of the endpoint in order to be able to issue meaningful queries. This task is particularly challenging for RDF as it is semi-structured data that tends to be heterogeneous and uses multiple vocabularies.
A common way to scrutinise the content of some SPARQL endpoint is via issuing a series of SPARQL queries1. These usually go from listing classes in the 1 For example see http://dallemang.typepad.com/ my_weblog/2008/08/rdf-as-self-describing-data.html and http://answers.semanticweb.com/questions/25696/ extract-ontology-schema-for-a-given-sparql-endpoint-data-set endpoint, to properties of each class, to sample instances of some classes, etc. Furthermore, to get an overview of the structure and relationships in the data, a more complex sequence of SPARQL queries are needed. We argue that this is a tedious, error-prone and time consuming task to be carried out manually. SPARQture, the tool described in this paper, automates this process.
SPARQture extracts and explicitly presents the schema used in a particular RDF data available behind some SPARQL endpoint. The tool issues a number of queries and then renders the information it collects using a number of visually rich and well-known information visualisation components. Furthermore, SPARQture supports user interaction to further explore the data. In the remainder of this paper, we provide more details about SPARQture (section 2) and discuss its t to the IESD Challenge by addressing the three key themes of the workshop: human factors, computational models and application domains (section 3). 2
Figure 1 shows the main interface of our tool, SPARQture. The main interface comprises two main panels: the tag cloud panel (top part) and the structure panel (bottom part). The tag cloud panel provides an overview of the data by displaying the types of entities in the endpoint. On the other hand, the structure panel allows getting detailed view of the structure of entities of a particular type. The structure is presented as a tree and can be further navigated by clicking on the nodes.
The user starts by entering the URL of a sparql endpoint or a VoID2 le
location. SPARQture presents a tag cloud of types in the endpoint where the size
of a tag re ects the number of instances of that type. By clicking on some type,
a set of sample resources are shown in the tag cloud panel and the structure of
some of these resources is plotted in the structure panel. It is worth mentioning
here that the root of the tree does not have to be a single resource but can
represent multiple resources. In the process of navigation, a node can represent
multiple values. By clicking on such node, SPARQture combines the structure
of the relations of these resources collectively, hence, supporting what can be
referred to as set-based browsing [
Finally, we summarise what we believe are the novel aspects of SPARQture: { Extracts and explicitly presents the schema used in the underlying RDF data. { Provides multiple visually-rich views over the data that allows exploring it at various levels of abstraction.
{ Can run on top of any endpoint that supports SPARQL 1.1. 3
3.1
SPARQture combines a number of familiar components to support presenting
and exploring information. Speci cally, tag clouds and tree visualisations are
used. Tag clouds utilise font size and colour as visual indicators of the
importance of an item as recommended in [
The choice of a tree instead of a graph to present data structure is based
on our belief that trees are easier to comprehend. By allowing duplicate nodes
in trees, graphs that contain cycles can still be represented as trees. A similar
decision was made in the Tabulator browser [
The core of SPARQture is an engine that translates user's interactions into a series of SPARQL queries. The results of the SPARQL queries are serialised as JSON objects that are rendered using JavaScript on the client side. We use Jquery TagCloud library3 to render tag clouds and reuse Javascript code from Open Re ne 4 project to render trees.
In the following we highlight a number of aspects of the computational model: SPARQL queries generation: We try to group SPARQL queries together to reduce the number of queries issued at the SPARQL endpoint and enhances the interface responsiveness at the same time. For example, the resource representing Dublin in DBpedia5 has 261 related resource (i.e., a triple with a URI object). Nevertheless, SPARQture issues only 8 queries to compute the tree structure with depth 3 rooted at Dublin. This is achieved via sending queries based on unique properties and grouping multiple properties in one query.
Content discovery: The entry point for the tool can be a SPARQL endpoint URL or a VoID le. SPARQture tries to locate the SPARQL endpoint location from the VoID le by searching for void:sparqlEndpoint property. We are currently working on extending this discovery feature to any resource URI by consulting the Endpoint Lookup Service6.
A few heuristics: We embedded a number of heuristics in the tool to select resources shown as sample resources or used as a root in the tree. 3.3
SPARQture is independent from the application domain. It can support any domain as long as an endpoint that supports SPARQL 1.1 is provided. 4
Extracting the schema of some semi-structured data has also been studied in the
context of XML data [
To conclude, we described SPARQture, a tool that aims at enhancing accessibility of SPARQL endpoints by providing a more welcoming interactive entry point. SPARQture uses tag clouds and trees to provide an overview of the data and supports interactive exploration of the data behind some SPARQL endpoint. Acknowledgements. Fadi Maali is funded by the Irish Research Council, Embark Postgraduate Scholarship Scheme.