Currency

  1. Age of data
  2. Specification of the modification date of statements
  3. Time since last modification

Age of data

The value regarding the KG creation date can be obtained from the VoID file or by executing a query on the SPARQL endpoint. In the VoID file we look for a triple having $dcterms:created$ as predicate. Instead the query for the endpoint should be of the type:

PREFIX dcterms: <http://purl.org/dc/terms/>
SELECT DISTINCT ?o
WHERE{?s dcterms:created ?o}
ORDER BY ASC(?o)
LIMIT 1

What this query does is retrieve all triples with predicate $dcterms:created$, then sorts the results in ascending order and takes the first value. This is because often multiple triples can be indicated with that predicate. To quantize the metric, if the creation date is indicated, then we assign 1 to this metric, 0 otherwise.

Specification of the modification date of statements

This value can also be obtained either from the file VoID or by executing a query. In the VoID file we look for the triple with predicate $dcterms:modified$, while on the SPARQL endpoint we execute the following query:

PREFIX dcterms: <http://purl.org/dc/terms/>
SELECT DISTINCT ?o
WHERE{?s dcterms:modified ?o}
ORDER BY ASC(?o)
LIMIT 1

In the opposite way to what happened for the creation date, here we sort the output in ascending order and take the first result. To quantize the metric, if the modification date is indicated, then we assign 1 to this metric, 0 otherwise.

Time since last modification

In this case we simply retrieve the modification date (with the previous query) and make the difference between the date on which we are performing the analysis and the date of the last modification. In this case we do not attribute a numerical value to the metric to possibly increase the score, because if we have the modification date we can certainly obtain this data, so the value of this metric depends on the modification date.

History of changes made

To calculate this data we need all the triples with $dcterm:modified$ predicate, which correspond to all the different modification dates. We use the same query used in the modification date metric, but we remove LIMIT 1. At this point, for each date we obtained from the query, we execute the following query:

PREFIX dcterms:<http://purl.org/dc/terms/>
SELECT DISTINCT (COUNT(?o) AS ?triples)
WHERE{
{?s dcterms:modified ?o}
FILTER regex(?o,’%s’)
}

In the regex function the %s parameter is set with the modification date for which we want to count the number of updated triples. In this case we do not attribute a numerical value to the metric to possibly increase the score, because if we have the modification date we can certainly obtain this data, so the value of this metric depends on the modification date.