Search by property
This page provides a simple browsing interface for finding entities described by a property and a named value. Other available search interfaces include the page property search, and the ask query builder.
List of results
- Querying the Web of Data with Graph Theory-based Techniques + (In this paper we present a novel graph the … In this paper we present a novel graph theory-based approach for improved performance and scalability of distributed SPARQL query processing. We propose several key techniques adopted in the distributed SPARQL query engine that we developed (GDS). First, an extended MST algorithm which supports arbitrary SPARQL queries and provides better performance. Second, a simplified cost model using run-time statistics that lows requirement of service descriptions but provides good cost estimations. Third, a combination of semi-join and bind-join along with local cache that reduce network tracffic. We also compare our approach with DARQ and FedX in terms of performance and scalability. The results suggest that graph theory-based approach using lightweight service descriptions can provide better performance and scalability over other approaches Although these results are encouraging, the potential of graph theory-based approach can be developed further. First, GDS applies MST-based algorithm to BGPs rather than the whole query. BGPs are optimized separately even they are from the same query (i.e. UNION and OPTIONAL queries). In the future we aim to work on representing the whole query as a single graph and therefore provide better optimization for queries having multiple BGPs. Second, GDS does not take advantage of FILTER in optimization, which would further improve performance. Third, accurate and detailed service descriptions matters. The more accurate statistics we have, the more optimal query plan we get. Currently collecting quality service descriptions are not feasible on a large scale since most SPARQL endpoints do not provide service descriptions. However, this situation is improving as more and more approaches coming up. For example, SPARQL 1.1 aggregation features enable us to collect service descriptions more efficiently, and VoiD encourages SPARQL endpoints to publish service descriptions as well. Fourth, aggregation features in the upcoming SPARQL 1.1 (e.g. BINDINGS) can also save much efforts of our approach. In addition to these improvements, we are planning to explore the co-reference issue in the Linked Data cloud. From the perspective of distributed SPARQL queries, this issue is getting worse as more data are published , and we plan to address this issue by using our Virtual Graph approach.issue by using our Virtual Graph approach.)
