Graph-based Metadata
This is an experimental area, discussing the feasibility and potential value of construction of a metadata architecture on top of a relatively simple graph-model.
Put simply, a metadata graph describes how things are related to other things. It doesn't go any further than that.
Syntactic Interoperability
A simple graph-model could provide an effective abstraction of a number of different metadata syntaxes, allowing metadata to be virtualised in a common, standard way.
In other words, a simple graph-model could be used as the basis for an abstract metadata syntax, providing a foundation for translation between various concrete metadata syntaxes.
On top of such an abstract syntax, application-specific constraints could be expressed, capturing the essential requirements of an "application profile".
Semantic Interoperability
An abstract syntax based on a simple graph-model also provides a firm foundation for a precise semantics.
A semantics tells you how to merge graphs, i.e. how to merge metadata from different sources.
A semantics also provides the foundation for layers of application-specific semantics. These layers then allow metadata to be "dumbed-down" gracefully, to whatever level is required.
Proof of Concept -- "Son of Dublin Core"
At DC-2007, AlistairMiles hacked up a proof of concept graph-based metadata architecture, building directly on the Resource Description Framework (RDF) abstract syntax. See the link below for more information:
-
http://purl.org/net/sodc/ -- Son of Dublin Core -- Encoding and Validating and Harvesting Graph-based (Meta)data