Effective And Efficient Keyword Query Using A Hybrid Graph On Semantic Data

In 2001, Tim Berners-Lee, the father of World Wide Web, and his colleaguespublished a popular science article"The Semantic Web"on journal of"ScientificAmerican". This article is the symbol of the birth of semantic web. SemanticWeb has developed more than ten years since then. The characteristics of se-mantic web are structured and semantic. Common semantic data is RDF data.Structured query language is the standard query language to access semanticdata, such as SPARQL query language. But the complex syntax of structuredquery hinder the promotion of semantic search. Ordinary users are accustomedto using a simple keyword queries. Although it is not as good as the structuredquery in the expression. However, because of its great advantage of convenient,it is widely used by existing search engines.Empowering users to access RDF data using keywords can relieve themfrom the steep learning curve of mastering a structured query language and un-derstanding complex and possibly fast evolving data schemas. In recent years,translating keywords into SPARQL queries has been widely studied. Approachesrelying on the original RDF graph (instance-based approaches) usually generateprecise query interpretations at the cost of a high processing time while those rely-ing on the summary graph extracted from RDF data (schema-based approaches)significantly speed up query interpretation disregarding the loss of accuracy. Inthis paper, we propose a novel approach based on a configurable hybrid graph,namely ICES, for the trade-o? between interpretation accuracy and e?ciency. Ascore function is then defined to assess the tradeo? between interpretation accu-racy and e?ciency for a given RDF graph, and thus helps guide the constructionof ICES iteratively. The derived hybrid graph is further used in the explorationfor the computation of top-k queries. The experiment is conducted on several widely-used data sets of di?erent sizes. The results show that we can achievesignificant e?ciency improvement with a limited accuracy drop compared withinstance-based approaches, while, our approach achieves promising accuracy gainat an affordable time cost compared with schema-based approaches.