This paper aims at introducing a generic architecture of a ontology extractor (Ontology Builder) from instances of a family of XML documents. This Ontology Builder comes from an ontology specification. After, it presents an instance of this architecture that process a family of XML documents, resulting, as output, a specification in XTM (XML Topic Maps). To describe the extraction of knowledge from XML documents to produce a XTM, a new XML language called XSTM (XML Specification for Topic Maps) is defined. The proposed architecture and language are illustrated by a real case study: the ontology specification connected to the authors and papers presented in a conference about XML (XATA) realized at University of Minho.

THIS PAPER IS WRITTEN IN PORTUGESE!