Interpretability

  1. No misinterpretation of missing values
  2. Atypical use of collections containers and reification

No misinterpretation of missing values

To count the number of nodes we use the following query:

SELECT (COUNT(?bnode) AS ?triples)
WHERE { ?bnode ?p ?o
FILTER (isBlank(?bnode))}

The query output is the number of blank nodes in the KG, and is used to understand the incidence of blank nodes compared to other resources. To quantize the metric, we first execute the following query that count the number of triples that do not have the property rdf:type:

SELECT (COUNT(?o) AS ?triples)
WHERE{?s ?p ?o.
FILTER(?p NOT IN (rdf:type))
}

Then we use the following formmula where $numBN$ is the output of the first query and $numDlc$ is the output of the second query:

\[m_{BN} = 1 - \frac{numBN}{numDlc}\]

Atypical use of collections containers and reification

In this case we check that there are no RDF data structures, in fact their use comes discouraged by W3C. To check their use we use the following query:

PREFIX rdf:<http://www.w3.org/1999/02/22-rdf-syntax-ns#>
PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
SELECT (COUNT(?s) AS ?triples)
WHERE{
{?s rdf:type rdf:List }
UNION
{?s rdf:type rdf:Statement}
UNION
{?s rdf:type rdf:Alt}
UNION
{?s rdf:type rdf:Bag}
UNION
{?s rdf:type rdf:Seq}
UNION
{?s rdf:type rdf:Container}
UNION
{?s rdf:subject ?o}
UNION
{?s rdf:predicate ?o}
UNION
{?s rdf:object ?o}
UNION
{?s rdfs:member ?o}
UNION
{?s rdf:first ?o}
UNION
{?s rdf:rest ?o}
UNION
{?s rdf:_’[0-9]+’}
}

If the query returns even just one result, it means we have a triple that declares the use of RDF structures and therefore we attribute a value of 1 to the data (i.e. they are used), 0 otherwise.