k-nearest keyword search in RDF graphs

Resource Description Framework (RDF) has been widely used as a W3C standard to describe the resource information in the Semantic Web. A standard SPARQL query over RDF data requires query issuers to fully understand the domain knowledge of the data. Because of this fact, SPARQL queries over RDF data are not flexible and it is difficult for non-experts to create queries without knowing the underlying data domain. Inspired by this problem, in this paper, we propose and tackle a novel and important query type, namely k-nearest keyword (k-NK) query, over a large RDF graph. Specifically, a k-NK query obtains k closest pairs of vertices, (vi, ui), in the RDF graph, that contain two given keywords q and w, respectively, such that ui is the nearest vertex of vi that contains the keyword w. To efficiently answer k-NK queries, we design effective pruning methods for RDF graphs both with and without schema, which can greatly reduce the query search space. Moreover, to facilitate our pruning strategies, we propose effective indexing mechanisms on RDF graphs with/without schema to enable fast k-NK query answering. Through extensive experiments, we demonstrate the efficiency and effectiveness of our proposed k-NK query processing approaches.