Efficiently Finding Paths Between Classes to Build a SPARQL Query for Life-Science Databases

Atsuko Yamaguchi, Kouji Kozaki, Kai Lenz, Hongyan Wu, Yasunori Yamamoto, Norio Kobayashi

Abstract: Many databases in life science are provided in Resource Description Framework (RDF) model with SPARQL Protocol and RDF Query Language (SPARQL) endpoints. However, it may be difficult for users who are not familiar with Semantic Web technologies to write a SPARQL query. Therefore, assisting users to build SPARQL queries is important task to expand the range of users of RDF databases. We developed a web application called SPARQL Builder (http://sparqlbuilder.org/) that enables users to access life-science RDF datasets by assisting them in writing SPARQL queries. One of the key technologies used in SPARQL Builder is to extract possible relationships in an RDF dataset between two classes of input and output data. We express such relationships by paths on a labeled graph called class graph representing class–predicate–class relations in a dataset. In addition, we present an efficient algorithm to compute all the possible paths between two classes on a class graph. To show the performance of the proposed algorithm, we compared our algorithm with a naive method using RDF datasets of various class sizes and confirmed that our algorithm runs much faster when the numbers of classes and relations are relatively large.