Data sources… exposed

In response to feedback from early beta testers, we have added the capability to view the sources of each fact within the portal. This is particularly helpful identifying the source for a specific value or relationship you believe to be incorrect.

The “view sources” link can be found at the bottom of every page that describes a capability/technique, person or organisation, and exposes the raw data within the portal, broken down by source.

Within the portal all information is held (and published) in accordance with Linked Data principles, using RDF as the data representation language. Without getting too technical, the main points to note are

  • all data is represented in “triples” – two concepts and a relationship between them
  • we use URIs (web addresses) to identify each concept
  • with Linked Data, to find out about a given concept all you need to do is visit it’s web address

Using the simple principles of associating concepts (be they real world entities such as people or places, or more abstract notions such as a topic or expertise) with well defined relationships allows us to form powerful graph structures.

Take the following model as an example…

Example RDF Graph

The orange rectangles represent concepts – people, organisations, areas of expertise – and each one would be represented by a web page describing all the information about that concept. Names, labels, email addresses, telephone numbers, etc are simply text values associated with a concept.

The key point to note is that the concepts are identified by their unique web address, not an ambiguous name. As more and more simple facts are added to the portal, the resulting graph based data representation becomes increasingly integrated as common occurrences map to the same concepts – that is to say, if more people are declared a having an expertise in a particular concept, then all of those people will have an association with the same concept node in the graph. (for a more complete representation of the graph, please see here).

Finally, within in the portal we have a system for mapping between equivalent concepts, as it would be rather idealistic to presume that every source will use the same terminology or identifiers. This also allows us to associate concepts from external sources, such as Freebase, with those within the portal. Any applicable equivalent terms (synonyms) are listed at the top of each “view source” page.