Toward RDF normalization

Abstract

Billions of RDF triples are currently available on the Web through the Linked Open Data cloud (e.g., DBpedia, LinkedGeoData and New York Times). Governments, universities as well as companies (e.g., BBC, CNN) are also producing huge collections of RDF triples and exchanging them through different serialization formats (e.g., RDF/XML, Turtle, N-Triple, etc.). However, RDF descriptions (i.e., graphs and serializations) are verbose in syntax, often contain redundancies, and could be generated differently even when describing the same resources, which would have a negative impact on their processing. Hence, we propose here an approach to clean and eliminate redundancies from such RDF descriptions as a means of transforming different descriptions of the same information into one representation, which can then be tuned, depending on the target application (information retrieval, compression, etc.). Experimental tests show significant improvements, namely in reducing RDF description loading time and file size.