RDF export

[simongray]

Currently, the system is bootstrapped from various existing data sources, but there is no official export.

In the exported version, the different input datasets will need to be separated entirely, necessitating a separation into separate graphs or some other means of data separation at the point of export.

Jena (SPARQL in general) has the concept of named graphs, however the way this is implemented in Jena seems to be basically as a union of separate graph objects. One major complication is the fact that the web frontend currently relies on inference to generate "missing" triples and this is done for a single "data" graph + a "schema" graph, so having multiple graphs will likely not work.

[simongray]

It should be possible to create

• a Dataset consisting of multiple named models
  • where each model of course has its own separate graph
    • and each of these models/graphs can be joined into a union model, e.g. using (.getUnionModel dataset) for querying
      • this union can then be the input for the inference model

[simongray]

I have successfully partitioned the input data into separate named models with separate graphs.

The bootstrap process returns the same data as before i.e. :dataset, :model, :graph. The main difference is that now the dataset contains the named models, while the model and the graph are so-called union graphs or—in the case of the inference model—an inference model based on this union model and the graph of the inference model.

It should be possible to simply write the named models inside the dataset to disk to get each individual dataset as an RDF export. The inference model can be written to disk to get the entire inferred dataset.