Processing select parts of an RDF graph with FREME

[fsasaki]

I am wondering how to achieve the following process using FREME: enrich selected parts of an RDF graph with additional information related to entities described in RDF triples with textual content. See this example
https://taskman.eionet.europa.eu/projects/linkeddata/wiki/DCAT-AP_example
Here you have triples like

<http://nobelprize.org/datasets/dcat#catalog>
a dcat:Catalog ;
dc:title "Nobel Media Dataset catalog"@en ..

only the object "Nobel Media Dataset catalog" should be processed in e-entity.

I understand that this may be a question for e-internationalisation. Hence putting @borriellom into the loop. But probably e-internationalisation cannot provide the answer, since it relies on a fixed syntax (of e.g. HTML or XLIFF input), and RDF can come with various syntaxes.

So thanks for the feedback of everybody in advance :)

[jnehring]

There is a possibility with the filter endpoint that is not yet implement but in discussion right now. You can remove everything from RDF graph that you do not need and then process only the sub graph.
Unfortunately the part of the RDF graph you dont want to enrich gets lost in this way.

[fsasaki]

Thanks, @jnehring . The issue may be that

<http://nobelprize.org/datasets/dcat#catalog>
a dcat:Catalog ;
dc:title "Nobel Media Dataset catalog"@en .

is not NIF, it is general RDF. Can general RDF be input to a FREME service?

[jnehring]

I think that using a (very complicated) SPARQL query you could transform the general RDF into NIF. The normal e-Services cannot process general RDF but there is no reason for the filter endpoint to forbid general RDF.

But this approach is very complicated. A lot easier to implement is the workflow to 1) Construct a SPARQL query that extracts the plain text on the client side and 2) sending the plaintext to FREME for enrichment.

[fsasaki]

I see. The issue is that above example is just one piece of plaintext. The actual metadata records like
https://taskman.eionet.europa.eu/projects/linkeddata/wiki/DCAT-AP_example
have a lot of plain text strings, so one would need to have potentially a lot of requests to the FREME enrichment services.