For research institutes, data libraries, and data archives, RDF data validation according to predefined constraints is a much sought-after feature, particularly as this is taken for granted in the XML world. Based on our work in the DCMI RDF Application Profiles Task Group and in cooperation with the W3C Data Shapes Working Group, we identified and published by today 81 types of constraints that are required by various stakeholders for data applications. In this paper, in collaboration with several domain experts we formulate 115 constraints on three different vocabularies (DDI-RDF, QB, and SKOS) and classify them according to (1) the severity of an occurring violation and (2) the complexity of the constraint expression in common constraint languages. We evaluate the data quality of 15,694 data sets (4.26 billion triples) of research data for the social, behavioral, and economic sciences obtained from 33 SPARQL endpoints. Based on the results, we formulate several findings to direct the further development of constraint languages.

1. Validating RDF Data Quality
using Constraints
to Direct the Development of Constraint Languages
Thomas Hartmann
Benjamin Zapilko, Joachim Wackerow, Kai Eckert
International Conference on Semantic Systems (ICSC 2016)

2. XML Validation

3. <!ELEMENT library (book+, author*)>
<!ELEMENT book (isbn, title, author-ref+)>
<!ATTLIST book
id ID #REQUIRED
>
<!ELEMENT author-ref EMPTY>
<!ATTLIST author-ref
id IDREF #REQUIRED
>
<!ELEMENT author (name)>
<!ATTLIST author
id ID #REQUIRED
>
<!ELEMENT isbn (#PCDATA)>
<!ELEMENT title (#PCDATA)>
<!ELEMENT name (#PCDATA)>

4. RDF Validation
Workshop

5. Working Groups on
RDF Validation
W3C Data Shapes Working Group
DCMI RDF Application Profiles Task Group

6. http://purl.org/net/rdf-validation
81 Types of Constraints
on RDF Data

7. Constraint
Languages

8. SPARQL Query Language for RDF
SELECT ?concept
WHERE {
?concept a [ rdfs:subClassOf* skos:Concept ] .
FILTER NOT EXISTS {
?concept ?p ?o .
FILTER ( ?p IN (
skos:related,
skos:relatedMatch,
skos:broader, ... ) ) . } }

9. SPARQL Inferencing Notation (SPIN)
# FILTER NOT EXISTS { ?book author ?person }
[ a sp:Filter ;
sp:expression [
a sp:notExists ;
sp:elements (
[ sp:subject [ sp:varName "book" ] ;
sp:predicate author ;
sp:object [ sp:varName "person" ]])]])

10. Web Ontology Language (OWL)
:Publication rdfs:subClassOf
[ a owl:Restriction ;
owl:onProperty :author ;
owl:allValuesFrom :Person ] .

11. Shape Expressions (ShEx)
:Publication {
( :isbn xsd:string, :title xsd:string )
|
( :issn xsd:string, :title xsd:string )}

12. Resource Shapes (ReSh)
:Computer-Science-Book
a oslc:ResourceShape ;
oslc:property [
oslc:propertyDefinition :subject ;
oslc:allowedValues [
oslc:allowedValue
"Computer Science" ,
"Informatics" ,
"Information Technology" ] ] .

13. [ a dsp:DescriptionTemplate ;
dsp:resourceClass :Science-Fiction-Book ;
dsp:statementTemplate [
dsp:property :subject ;
dsp:nonLiteralConstraint [
dsp:valueClass skos:Concept ;
dsp:valueURI
:Science-Fiction, :Sci-Fi, :SF ;
dsp:vocabularyEncodingScheme
:Science-Fiction-Book-Subjects ; ] ] .
Description Set Profiles (DSP)

14. Shapes Constraint Language
(SHACL)
:BookShape
a sh:Shape ;
sh:scopeClass :Book ;
sh:property [
sh:predicate :author ;
sh:valueShape :PersonShape ;
sh:minCount 1 ; ] .

15. http://purl.org/net/rdfval-demo
RDF Validation
Environment

17. Constraint Types
Classification
1. RDFS/OWL Based
2. Constraint Language Based
3. SPARQL Based

18. RDFS/OWL Based
:Publication rdfs:subClassOf
[ a owl:Restriction ;
owl:onProperty :author ;
owl:allValuesFrom :Person ] .

19. Constraint Language Based
:Publication {
( :isbn xsd:string, :title xsd:string )
|
( :issn xsd:string, :title xsd:string )}

20. SPARQL Based
SELECT ?concept
WHERE {
?concept a [ rdfs:subClassOf* skos:Concept ] .
FILTER NOT EXISTS {
?concept ?p ?o .
FILTER ( ?p IN (
skos:related,
skos:relatedMatch,
skos:broader, ... ) ) . } }

21. Constraints
Classification
1. Informational
2. Warning
3. Error

22. Evaluation Setup
• 115 constraints from vocabularies and
experts
• constraints classified and implemented
• on 3 vocabularies in the SBE sciences
– well-established vocabularies (QB, SKOS)
– vocabulary under development (DDI-RDF)

23. Validated Data Sets
Vocabulary Data Sets Triples
QB 9,990 3,775,983,610
SKOS 4,178 477,737,281
DDI-RDF 1,526 9,673,055
Total 15,694 4.26 billion
33 SPARQL Endpoints

24. Finding 1
C [%] CV [%]
SPARQL 63.2 78.2
CL 34.7 21.8
RDFS/OWL 35.6 21.8
C (constraints), CV (constraint violations)

25. Finding 2
C [%] CV [%]
SPARQL 63.2 78.2
CL 34.7 21.8
RDFS/OWL 35.6 21.8
C (constraints), CV (constraint violations)

26. Finding 3
C [%] CV [%]
Info 42.3 31.3
Warning 18.7 62.7
Error 39.0 6.1
C (constraints), CV (constraint violations)

27. Limitations
> 3 Vocabularies
> 1 Domain