(a) Area of occupation

(b) Size of graph (Number of triples)

(c) Size of ontology (Number of classes)

(d) Distinct # of properties in the graph

(e) Number of generated shapes

(f) Scope of the generated shapes

We see that researchers in academia and data scientists in industry require to define shapes from fairly large knowledge graphs with complex structures. Yet, there exists strong limitations of existing shapes extraction approaches in effectively supporting the real user needs while the common practice is that of generating shapes manually.

Our findings revealed that existing shapes extraction tools and approaches are capable of handling only relatively small knowledge graphs, they are unable to extract all the required shapes constraints, and are prone to output spurious or erroneous shapes. Read our paper for more details