Linked Data [1] is Semantic Web data that emphasises the graph of relations between resources while recognising that the data comes from Semantic Web documents that need to be retrievable using standard conventions. Domain-specific applications can crawl Linked Data to pull together and display information from various sources. However, there is a need for generic Linked Data browsers as well, as they help data producers to check what they publish and data consumers to check what is available.
One generic Linked Data browser is Tabulator [2], but it does not provide a graph view. Karger and schraefel provide insights on why graph views often are not the best views of Semantic Web data [4], but they acknowledge that graph views have a high "cool factor" and a niche. We consider a case that lies in this niche: we want to show the graph data as directly as possible. We apply a Free and Open Source Software rich desktop application, Fenfire 1 , and its graph view to Linked Data browsing because unlike other browsers, this is a visually appealing, engaging and interactive demonstration of the Semantic Web's capabilities. A graph view is a good way to explore a web of information, and it is close to the nature of Linked Data as a heterogeneous, web-like environment with little high-level structure.
In the following, we demonstrate how the Fenfire applica- tion provides a useful Linked Data browsing experience, go through an example, describe the implementation and conclude with some future directions.
The Fenfire application is a generic RDF browser and editor with features useful for Linked Data browsing. The user interface employs the conventional graph representation of the RDF data model. To make the visualisations scalable in the number of nodes in the graph and to focus on one thing at a time, only one central node and its surroundings are displayed concurrently. It is possible to switch between two views implemented based on the concept: a simple list view of objects associated with the focused subject (say, a container), and the generic graph view from Fentwine [3].
A browsing session starts from some URI which is retrieved for a document with RDF data. This URI will be the initial focus unless the document has a foaf:primaryTopic defined, in which case the primary topic URI will be chosen as the initial focus.
For the surroundings, the generic graph view shows to the left of the focus all triples that have the focus as an object, and to the right all triples that have the focus as a subject. Each triple is shown as a predicate connecting the subject to the object. This view is applied recursively to each displayed node until, with distance, the graph fades away to the background. If the node has an rdfs:label, it is displayed instead of the URI.
Graph navigation can be done entirely via keyboard by rotating the surrounding nodes around the centre and moving focus to the node immediately to the left or right of the centre. While there is limited space for the surrounding nodes, all nodes can be navigated to via the rotation.
To enable browsing of Linked Data, Fenfire dereferences the URI of the focused node and retrieves any rdfs:seeAlso related to this node whenever instructed to do so. For example, in Figure 1 a user has loaded the FOAF profile of one person, followed a foaf:knows link to another person and retrieved the FOAF profile of this person.
As an important alleviation of incompletely linked data, Fenfire adds triples asserting that the graph retrieved contains information about all of its disconnected components.
We target two audiences primarily:
• Semantic Web researchers, application developers and data producers need to explore available data on the level of individual triples. Fenfire provides a convenient alternative to manually downloading graph documents, reading the serialisation formats, and trying to match URIs to discover the links in the data.
• People who want to learn about or demonstrate the Semantic Web and what data is available benefit from a visual presentation that truthfully shows the networked nature of the data. Here it is highly advantageous that Linked Data documents that follow the guidelines include an rdfs:label for each node.
Fenfire has its origins in ZigZag [6], which is a system for managing interlinked, distributed data and is completely independent of the World Wide Web standards. Fentwine [3] incorporated Semantic Web standards and became a graphview RDF browser, as did its sibling BuoyOING [5] (Buoy-Oriented Interface, Next Generation) user interface, which adds spatial navigation and concentrates on it.
Fenfire is implemented in the Haskell programming language to achieve high programmer productivity while meeting performance requirements for real-time animation and for browsing large graphs. It uses the Raptor library of the Redland project for RDF parsing, the GTK library for the standard graphical user interface elements and the Cairo library for the animated, vector-graphics based visualisations.
A major part of the architecture is the key-frame and identitybased visualisation and animation system that provides the user continuous feedback on how the navigation and the switching between alternative views change what is visible.
We applied Fenfire to the task of browsing Linked Data with a graph view and highlighted the use cases of exploring, learning about and demonstrating Semantic Web data. Some Linked Data browsers can edit the data as well. Fenfire can edit RDF graphs, but we need to implement remote publishing interfaces in addition to the current feature of saving to local files.
Information display and navigation can be enhanced if there is knowledge about the user interaction requirements for a specific domain. Thus, Fenfire should have some automatic ontology-awareness and more should be configurable with settings and plug-ins.
The implementation of Fenfire in Haskell was initially developed by Benja Fallenstein and Tuukka Hastrup.