Search code examples
sparqlskos

Find URI based on literal variable using FILTER and REGEX

I need to find URIs of skos:Concept using skos:prefLabel (literal) from other skos:Concept. Here is my query:

SELECT ?variableURI ?variablePref ?entityPrefRegex ?entityURI ?entityPref WHERE {
    ?variableURI skos:prefLabel ?variablePref .
    FILTER(REGEX(?variablePref,"^Dissolved .* in surface water"))
    BIND(REPLACE(?variablePref,"^Dissolved (.*) concentration in surface water", "$1") AS ?entityPrefRegex).
    ?entityURI skos:prefLabel ?entityPref .
    FILTER(REGEX(?entityPref,?entityPrefRegex,"i")) 
}

My problem is that the filtering part return no result and I don't understand why.

Here are sample variables I'm trying to link my entities

variableURI variablePref entityPrefRegex
<:c_7e508e0e> "Dissolved aluminium concentration in surface water"@en "aluminium"@en
<:c_b5dec35c> "Dissolved arsenic concentration in surface water"@en "arsenic"@en
<:c_bc765ffd> "Dissolved boron concentration in surface water"@en "boron"@en
<:c_4ce4d2c7> "Dissolved caesium concentration in surface water"@en "caesium"@en

And the corresponding entities. As you can see the literal are identical except for the capital letter.

entityURI entityPref
<:c_d57d0742> "Aluminium"@en
<:c_d57d077> "Arsenic"@en
<:c_d57d0728> "Boron"@en
<:c_d57d0745> "Caesium"@en
Solution

  • The pattern (second) argument to REGEX is a "simple literal" (a literal "without language tag or datatype IRI"). In this case, it looks like you are using ?entityPref values that have the @en language tag:

    FILTER(REGEX(?entityPrefRegex,?entityPref,"i"))

    Try instead casting the pattern to a plain string:

    FILTER(REGEX(?entityPrefRegex,STR(?entityPref),"i"))