140 likes | 274 Views
Semantic Web Instance Data Evaluation. Jiao Tao September 4, 2008. Instance Data Evaluation. Semantic Web data includes: schema statements, e.g. Wine instance data, e.g. (W rdf:type Wine)
E N D
Semantic Web Instance Data Evaluation Jiao Tao September 4, 2008
Instance Data Evaluation • Semantic Web data includes: • schema statements, e.g. Wine • instance data, e.g. (W rdf:type Wine) • Instance data evaluation checks whether the instance data (D) comply with the schema statements in its referenced ontologies (O)
Issues in Instance Data • Syntax errors Can D be parsed with certain RDF parsers? • Logical inconsistencies Is D+O consistent? • Potential (Integrity) issues (use CWA) Does D conforms to the schema in O? Ignored by most semantic web evaluation tools, because they use OWA!
Potential Issues • Unexpected individual type (UIT) • Redundant individual type (RIT) • Non-specific individual type (NSIT) • Missing property value (MPV) • Excessive property value (EPV)
Unexpected Individual Type • Description: no rdfs:subClassOf relation between the individual’s declared type in D and the types expected by O • Example:
Redundant Individual Type • Description: an individual is declared to be an instance of a class and its super-class in D at same time • Example:
Non-specific Individual Type • Description: an individual is declared to be an instance of a class which is a non-leaf node in the class hierarchy, no leaf node are declared to be its type. • Example:
Missing Property Value • Description: let m=|(i p x)| in D, assume O requiring i has at least n values for p, if m<n, then MPV issue would occur. • Example:
Excessive Property Value • Description: let m=|(i p x)| in D, assume O requiring i has at most n values for p, if m>n, then EPV issue would occur. • Example:
Issue Detection – SPARQL SELECT ?i ?p ?x ?c1 ?c2 FROM … WHERE { GRAPH D { ?i ?p ?x .} GRAPH DC { ?i rdf:type ?c1. ?p rdfs:domain ?c2. OPTIONAL {?c1 rdfs:subClassOf ?c3 . ?c3 rdfs:subClassOf ?c2 .} FILTER( !BOUND(?c3) ) OPTIONAL {?c2 rdfs:subClassOf ?c4 . ?c4 rdfs:subClassOf ?c1. } FILTER( !BOUND(?c4) ) }} //c1(Winery) is not compatible with expected type c2 (Wine)
Characterizing the Potential Issues Using ADLs • ADLs extends DLs with two modal operators: • K: minimal knowledge • A: assumptions • Most potential issues could be formalized with ADLs. However there is no reasoner can handle the reasoning of autoepistemic extensions of OWL. • SPARQL-based approach is a sound solution to the potential issues which are characterized using ADLs.
Future Work • To apply the instance data evaluation approach into TW wiki, e.g. whether a page conforms to a template?