Metacat

Extending the model

___________________

The proposed system for asserting and indexing annotations can easily be extended. For practical reasons, we do want to codify a preferred mechanism for expressing 

observation measurements and binding them to their data table attributes. But because the model is essentially just a collection of triples, and the mechanism that indexes those

triples is configured with custom SPARQL queries, we can accommodate additional statements about data objects and packages in the future.

One such semantic extension involves a provenance graph for derived data products. For detailed information on that endeavor, see the ore-model-expansion section.

Another area for extension uses ORCIDs to give attribution to the appropriate author/creator. This is expressed in the model using prov:wasAttributedTo and 

could be readily indexed into a dynamic SOLR field like ``creator_sm``. But until these ORCIDs are more widely adopted, it may be difficult to provide effective querying based on this field.

It would also require authors to actively assert that their ORCIDs are associated certain data packages and objects; perhaps using tools that we currently do not have implemented. 

Annotation serializations

______________________________

Our initial serialization technique for semantic annotations is to have a distinct file for the model. We have been using RDF/XML, but other syntaxes will likely be supported out of the box 

because we are using the Jena library for model parsing.

Other methods for serializing the model we have considered and may support in the future include:

	* ``ORE`` - included as additional triples in our current ORE resource map packaging serializations

	* ``RDFa`` - annotations embedded directly within the science metadata

	* ``triplestore`` - triples written directly to a triple store endpoint using an API

Annotation permissions

______________________________

Because annotations ussually assert facts (or opinions) about _other_ objects, we will allow these assertions to be indexed only if the rights holder for the RDF model has the same rightsholder

priveledges on the target object.

This will prevent both malicious and accidental assertions about objects by other parties who should not be influencing how the object is documented or interpreted. 

Unfortunately, this also prevents interested, non-rights holder parties from asserting valuable statements about research data in the system. 

Ideally, we will accommodate third-party annotations and expose them for use in discovery and integration so long as they are effectively labeled (e.g., "alternative annotaiton", "automated annotation", etc...).