Checking compliance of semantic web applications with RDFS-semantics

Web applications calling themselves "Semantic" count in the hundreds, yet there is no clear definition about what this qualification should stand for. Semantic web applications may range from those just using Resource Description Framework (RDF) as a data interchange format, to those taking all RDF-S consequences into account. Moreover, even with the simplest semantics, RDF data may contain blank nodes, which should be treated appropriately as existential variables-but might not. In this paper, we propose a general framework of yes/no experiments whose result tell, in a black-box fashion, how "Semantic" can be considered a given Web application. Our experiments measure the sensitivity of the application to syntactic variations of data which are equivalent under RDF-S-semantics. We show how our experiments can be run on a real application, namely, Rapid Miner with LODEx-tension. We show how RapidMiner passes most of the tests (but for blank nodes), but highlight a weakness of the workflow repository-retrieval-exploitation that may make the application fail in some cases.