Availability

  1. Accessibility of the SPARQL endpoint
  2. Accessibility of the RDF dump
  3. Derefereaceability of the URI

Accessibility of the SPARQL endpoint

First of we need to check that it is present for the KG we are considering. The SPARQL endpoint link can be recovered in three different ways:

  1. The first (easiest) is to analyze the metadata and search for the resource with the tag in the resources field api/sparql or whose key is sparql.
  2. Just in case this way is unable to recover it, then we proceed to search inside the VoID file if it is available. In this case we go in search of the triple having void:sparqlEndpoint as predicate.
  3. The third and final method involves retrieving the URL of the dataset and adding /sparql to it, since this is the position that is usually given to the SPARQL endpoint on the server.

If none of these three methods leads to having a SPARQL endpoint, then it is declared that this KG does not have one and is given to it assigned value -1. In case we manage to recover the link instead, a simple query is run on the SPARQL endpoint to test whether it is online or offline.

SELECT ?s
WHERE {?s ?p ?o .}
LIMIT 1

If we get the triple correctly in the answer, then the SPARQL endpoint is declared active and a is assigned value equal to 1. Otherwise, if errors occur during the query, or no result is returned, then it is declared offline and given value 0.

Accessibility of the RDF dump

To check for the presence of the RDF dump we have three possible approaches:

  1. We can analyze the metadata and check if in the resources field there are one or more resources with one of the following tags: application/rdf+xml, text/turtle, application/x-ntriples, application/x-nquads, text/n3, rdf,text/rdf+n3, rdf/turtle.
  2. Another method is to check inside the VoID file (if available). In this case we search for the triple having void:dataDump as its predicate.
  3. Finally, the last way involves execution of a query on the SPARQL endpoint (if online). The query executed is as follows: 
    PREFIX void: <http://rdfs.org/ns/void#>
SELECT DISTINCT ?o
WHERE
{?s void:dataDump ?o}

Once the dump link has been retrieved, a simple HEAD request is made on the URL, to check whether it is online or not. If it is online, a value of 1 is given to the data, if offline 0 and if not present -1.

Derefereaceability of the URI

5000 triples (which contain URIs) are randomly retrieved with this query:

SELECT DISTINCT ?s
WHERE {
?s ?p ?o
FILTER(isIRI(?s))
}
ORDER BY RAND()
LIMIT 5000

Then a GET request is performed for each of them, specifying it in the header application/rdf+xml as accepted format. If it is returned status code 200 then the resource is available, otherwise it comes declared as unreachable. At the end of the test on all 5000 triple the following formula is applied to quantize the data, where $U_g$ indicates the set of URIs that are tested:

\[m_{def} = \frac{|Dereferencable(U_g)|}{|U_g|}\]