Ontology-driven search applications use ontological concepts either to index documents or to guide and understand the users. Since ontologies by nature are domain-dependent and application-independent, though, there is no guarantee that their concepts are efficient in categorizing and retrieving information from a specific document index. This paper explains the idea of ontological profiles, which is an ontology adapted to the actual language used in a document collection or among the users. A method for constructing ontological profiles from petroleum documents is presented, as well as a search application that makes use of profiles to interpret users queries. Testing on real documents with a 20,000 concepts petroleum ontology reveals that the approach is useful in situations where recall is more critical than precision.