Data Mashup is a special class of applications mashup that combines data elements from multiple data sources (that are often ex- ported as data web services) to respond to transient business needs on the y. In this paper, we propose a semantic model for data services along with a declarative approach for creating data mashups without any programming involved. Given a query formulated over domain on- tologies, and a set of preferences modeled using the fuzzy set theory, our approach selects the relevant data services based on their semantic modeling using an RDF query rewriting algorithm. Selected services are then orchestrated using a ranking-aware algebra to rank their returned results based on users preferences at the data mashup execution time.