We propose here a novel approach for extracting hidden semantic from semi-structured data resources and transformed to RDF triples, to be queried through semantic query languages. Unlike existing approaches that only explore the structure or use ontologies, we present a system that allows us to utilize all available information. Our approach constructs a semantically related data from XML represented in RDF via “semantic network” described here. These data could be queried by SPARQL/XQUERY or both simultaneously in multimodal way to perform a semantic search on a set of multimedia news resources.