Semantic search on summarized RDF triples

Abstract

Information is for the most part found inside databases of some kind. Coordinating these information would give advantages to the associations that claim these information. RDF is a general recommendation dialect for the Web, binding together information from different sources. SPARQL, a query language for RDF, can join information from various databases. Querying a huge RDF informational collection is to a great degree time consuming. Submitting a SPARQL query to a more promising subgraph will increase speed and reduce search space. For this, RDF summarization is done. Existing systems either use graph-based techniques for summarizing RDF or divide RDF triples simply based on its elements. RDF though possessing a graph like structure, cannot be expected to have every features of graph structure. And simple partitioning based on triple elements is also inefficient. In this paper, RDF dataset is first partitioned based on predicate similarity. These partitions are clustered based on semantic relatedness between predicates, so that more similar triples come in a single cluster. The RDF cluster graphs thus obtained are stored in Jena Tuple DataBase(TDB) as named graphs. SPARQL querying, is done on this named graph collection. A list of models that the SPARQL query require is obtained from index and querying is done on this union of graphs. The proposed algorithm is faster as search space is reduced and is also scalable.