Exposing heterogeneous data as RDF knowledge graphs is usually done with declarative mapping languages. However, the creation of mapping rules is an arduous task, which in the case of relational databases can be mitigated with the automatic extraction of mapping rules. For property graphs, it is convenient to map them to RDF-star which provides quoted triples as a more natural representation for edge properties. Nevertheless, the automatic extraction of declarative mappings from property graphs to RDF-star remains unstudied. In this paper, we address this problem using the recently proposed RML-star mapping language. We implement and validate the approach with the LDBC Social Network Benchmark.