RelFinder, a new Adobe Flex-based application, has been announced by a collection of researchers from the University of Stuttgart, the Carlos III University of Madrid and the University of Duisburg-Essen. Additional contributions have been made by individuals from the University of Leipzig and AKSW.

The application can be configured to point to arbitrary RDF datasets exposed through a SPARQL endpoint, but uses several standard data sets out of the box. The default sources include:

• DBpedia
• An endpoint to the Linking Open Data Project
• The Linked Movie Data Base
• DBTune
• Linking Open Drug Data (LODD)
• DBLP Bibliography Database

Users specify two or more concepts through the user interface to explore. When the "Find Relations" button is pressed the application starts issuing SPARQL queries and builds up a graph of found relationships. The automatic layout engine spatializes the results based on what has been found.

If intermediate object nodes are selected, a red line is drawn to highlight the relationships into and out of these objects. If an object node is pulled away from its automatic layout location, it can be pinned down with surrounding relationships restructured accordingly.

Users are able to filter the results based on the class of the intermediate objects, the types of relationships, the length of the connections, etc.

Several sample queries exist including looking for connections between "Albert Einstein" and "Kurt Gödel", "OWL", "RDF" and "XML" or "Kill Bill" and "Pulp Fiction". As you can see, you can easily convert an existing query into a sharable link.

You can download a screencast to see the application in action or watch it via streaming media on the main site. There are also several papers that describe the SPARQL queries and disambiguation processes used to find the relationships.