Abstract : The semantic web relies on ontologies representing domains through their main concepts and the relations between them. Such a domain knowledge is the keystone to represent the semantic contents of web resources and services in metadata associated to them. These metadata then enable us to search for information based on the semantics of web resources rather than their syntactic forms. However, in the context of the semantic web there are many possibilities of executing queries that would not retrieve any resource. The viewpoints of the designers of ontologies, of the designers of annotations and of the users performing a Web search may not completely match. The user may not completely share or understand the viewpoints of the designers and this mismatch may lead to missed answers. Approximate query processing is then of prime importance for efficiently searching the Semantic Web. In this paper we present the Corese ontology-based search engine we have developped to handle RDF(S) and OWL Lite metadata. We present its theoretical foundation, its query language, and we stress its ability to process approximate queries.