In the last decade, the field of Big Data Analytics has become increasingly important in both the academic and the business communities. Typically, data are mostly structured, collected by different actors through various heterogeneous and distributed information sources, and stored and managed often directly in XML. In order to enable large volume of data to be described in such a way that their meaning can be exploited by machines and, thus, semantic queries and automatic inferential procedures can be enabled, this paper presents an automatic method to derive OWL ontologies from XML schemas. The main contribution of this method relies on the possibility of producing a target ontology starting from multiple XML schemas, by discriminating between domain and cross-domain entities and, contextually, simplifying the overall structure of the final ontology generated, i.e. by eliminating not-used cross-domain entities. This method has been applied to a concrete application case in the healthcare domain, with the goal of generating an ontological model from the XML schemas implementing the HL7 Version 3 Clinical Document Architecture Release 2.