The widespread explosion of Web accessible resources has lead to a new challenge of locating all relevant resources and identifying the best ones to answer a query. This challenge has to address the difficult task of ranking the resources based on user needs, as well as the more expensive computational task of determining all the solutions to answer a query. In this paper, we define a Top K problem for query rewriting on the Semantic Web. We first introduce a three level data model composed of the ontology level, the physical level of the physical resources, and the data level composed of the entries in the different resources. We present a query language for Top K navigational queries over ontology concepts. We then sketch the outline of an efficient search algorithm to compute an approximation of the Top K rewriting options to produce source paths among the physical resources. We briefly discuss the results of an experimental study