Flexible querying techniques can be used to enhance users' access to heterogeneous data sets, such as Linked Open Data. This paper extends SPARQL 1.1 with approximation and relaxation operators that can be applied to regular expressions for querying property paths in order to find more answers than would be returned by the exact form of a user query. We specify the semantics of the extended language and we consider the complexity of query answering with the new operators, showing that both data and query complexity are not impacted by our extensions. We present a query evaluation algorithm that returns results incrementally according to their ``distance'' from the original query. We have implemented this algorithm and have conducted preliminary trials over the YAGO SPARQL endpoint and the Lehigh University Benchmark, showing promising performance for the language extensions.