Keyword Query For RDF Data Based On Query Translation

With the continuous development and application of the Semantic Web technology,we are inundated with a significant amount of rapidly growing RDF data which can be read and processed by machines.Keyword search for RDF data has become a hot topic of the semantic Web research.Both end users and application systems have querying demands for the RDF data.However,the standard query language for RDF data,SPARQL,is too complicated for ordinary users to master its syntax and semantics and it is too difficult to master the schema information of the RDF data.Therefore,we present a keyword search approach for RDF data by translating keyword search into SPARQL queries.Furthermore,we propose a two-phase optimization method for SPARQL queries produced so as to improve the query efficiency.The main contributions of our work are as follows:(1)We present a keyword search approach for RDF data based on a-condensed entity relationship summary graph.From the large scale RDF data,we extract the entities and relationships between these entities.In order to facilitate the query translation,the type of the entity is encapsulated in the entity node.Thus,a condensed entity relationship summary is built.By using the bidirectional search algorithm,we can find top-k subgraphs collecting all the keyword elements over the entity relationship summary.The top-k subgraphs corespond relationships of query variables.Then,we transform these subgraphs into SPARQL queries.Finally,we execute these SPARQL queries by existing excellent SPARQL search engines.(2)We present a keyword search approach for RDF data based on an entity relationship summary.By summarizing relationships between entities of RDF data,we define an entity type relationship summary.From the perspective of object query SPARQL query,during contructing the summary,we employ the property paths of SPARQL 1.1 including the predicate path operator,optional path operator the sequence path operator "/" and etc.The summary index not only makes the translation from the keyword query to the SPARQL query easier,more convenient and more efficient,but also makes up for the limits of existing index for query translation,because it completely summarize all entity type relationships of RDF data.Eventually,on the summary,we find top-k subgraphs containing all keyword entity type relationships transform these subgraphs into SPARQL queries and execute the generated SPARQL queries by existing SPARQL search engines.(3)We present a keyword search approach for RDF data using multiple indexes built.An inter-entity relationship summary is constructed by distilling all the inter-entity relationships of RDF data graph.On the summary,we draw circles around each vertex with a given radius r and in the circles we build the shortest property path index,the shortest distance index and the r-neighborhoods index.These indexes enable the transition of keyword queries to SPARQL queries.Although there exists storage overhead of indexes,through exchanging space for time,we greatly improve the query efficiency.Finally,the produced queries are implemented by existing SPARQL search engines.(4)We propose a two-phase query optimization method to optimize the produced queries by our approach.In the first phase,we group the join patterns sharing a join variable into building blocks of the query plan since executing them first provides opportunities to reduce the size of intermediate results generated.In the second phase,we further filter the intermediate results by the longest property path index.We employ both the property paths of RDF data graph and the selectivity of the triple pattern by which we greatly reduce the intermediate results produced during joining triple patterns so as to improve query quality.