SPARQL processing over the linked open data with automatic endpoint detection

Abstract

The LOD Cloud is a collection of the available datasets of the Semantic Web. The individual datasets typically store information about one specific area. These datasets can be queried using the SPARQL query language. Since the SPARQL 1.1 standard, it is possible to query several remote datasets with one query, and to combine the received data. This can be done with the SERVICE clause. SERVICE clauses generate join operations, and each join operation increases the load of the endpoint, that is why some endpoints do not allow at all to run queries using the SERVICE clause on their web interface. Some endpoints use SPARQL 1.0 and they do not know the SERVICE clause. There are endpoints which estimate the execution time of the query, and if that reaches a limit, then the endpoint refuses to execute it. Our aim is to create a system, which is able to execute queries over the LOD Cloud, without explicitly mentioning the necessary endpoints in the SERVICE clauses. Instead, the proposed system automatically recognizes the endpoints based on conditions of the query.