Temporal Extension for SPARQL

Abstract

A triplestore is a database for storing and retrieving Resource Description Framework (RDF) data. RDF is used to model the data on the Web, enabling Web-wide data merging and sharing. RDF is becoming more and more important as the need for a Semantic Web grows. To extract information from RDF data, W3C recommended a query language called SPARQL. However, SPARQL queries are not able to retrieve temporal data. It is desirable that a database maintains data together with temporal information about the data, e.g. when the data were created and changed. There are many applications that need both the current state of a database, as well as the history of the data. In order to handle the history of data, a number of temporal extensions to relational databases and triplestores have been proposed. The lightweight triplestore PDStore, developed at the University of Auckland, adds a transaction time and a change type to each triple, and can thus be used to answer temporal queries. This thesis presents the definition of a temporal extension of SPARQL, called Temporal SPARQL, which fills the gap that exists currently for temporal data in triplestores. Furthermore, the thesis presents the design and implementation of a Temporal SPARQL query engine. The temporal SPARQL engine is able to perform complex temporal queries, as Temporal SPARQL has the ability to handle time slices and time intervals. Its completeness is also discussed in the thesis. A performance comparison with another triplestore has been conducted, and the results are discussed. The implementation presented here is slower than the third-party engine, which does not come as a surprise as the third-party engine has undergone several years of development. However, the comparison highlights directions that can be taken to improve the performance in the future.