Answering SPARQL queries using views

Views are used to optimize queries and to integrate data in Databases. The data integration schema is composed of terms, they are used to pose queries to the integration system, and to describe sources data. When the data descriptions are SPARQL conjunctive queries, their number and the complexity of answering queries using them may be very high. In order to keep query answering cost low, and increase the usability of views to answer queries, we make the assumption of the simplest form of replication among services, and use triple pattern views as the Linked Data Fragments and Col-graph approaches have recently done. We propose two approaches: the SemLAV approach to integrate heterogeneous data using Local-as-View mappings to describe Linked Data and Deep Web sources, and the FEDRA approach to process queries against federations of SPARQL endpoints with replicated fragments.