Provenance Management for SPARQL Updates

During the last few years we have witnessed an explosion in the publication of data in the Web, mainly in the form of Linked Data. Scienti c, corporate or even governmental data are made available for open access and used by applications, individual users and communities. Given the increasing amount and the heterogeneity of this data, it is of crucial importance to be able to track its provenance. Recording the provenance can help us to e ectively support trustworthiness, accountability and repeatability in the Web of Data. A number of models have already been proposed to capture the provenance information of query results; most of them considering RDF or relational data. On the contrary, despite its importance, little research has been conducted in the case of updates and especially of SPARQL updates. In this thesis, we propose a new provenance model that borrows from both how and where data provenance models, and is suitable for capturing the triple and attribute level provenance of SPARQL update results. To the best of our knowledge, this is the rst model that deals with the provenance of SPARQL updates using algebraic provenance expressions, in the spirit of the well-established model of provenance semirings. On the algorithmic side, we introduce an algorithm that records the provenance of SPARQL update results in terms of the proposed model and a reconstruction algorithm that uses the provenance of a quadruple to identify a SPARQL update that is provably compatible to the original one. A SPARQL update is compatible to another if they di er only in the variables names that they employ and the rst update contains a genuine subset of the unions that appear in the second one. The latter algorithm is a necessary complement in order to fully describe the provenance management, as it shows the determinant role of provenance information in the persistence of SPARQL update results.