Towards Using NOSQL Database Systems for Semantic Web Data

Abstract

The Semantic Web allows data to be shared and reused through a mechanism of URIs and resource description. RDF, as a standard for describing resources and data, models information in the form of subject- predicate-object triples. This model can be viewed as a graph; subjects and objects are vertices or nodes and are connected through predicates as edges of the graph. Relational database systems are not e cient in processing Se- mantic Web data. With the massive increase in graph data, attributed to the growth of social networks, many native graph database systems have surfaced to replace Relational databases. Neo4j is the leading and very scalable graph database system that can store up to millions of nodes and relationships. Storage of Semantic Web data in Neo4J requires mapping of RDF constructs to Neo4j. Henceforth, retrieval requires mapping of the SPARQL query lan- guage for Semantic Web, with the Cypher query language for Neo4j. Logical inference or reasoning is a fundamental building block of the Semantic Web. A complementary standard in the SemanticWeb architecture, known as RDF Schema or RDF/S for short, provides a set of inference rules that de ne the mechanism for discovering and generating new relationships based on exist- xii LIST OF FIGURES xiii ing data. An e cient and scalable implementation of RDF/S inference rules is also missing in most of the graph database systems. This paper contributes three pieces of the jigsaw to achieve a scalable RDF storage system. First a mechanism for mapping RDF constructs with Neo4J is outlined. Secondly, SPARQL query language is mapped with Cypher for e cient retrieval. And nally, RDF/S inference rules are implemented to realize logical inference capability.