The nature of information sources together with the multiple document description alternatives offered by these sources are permanently changing. This evolvement was accelerated by the expansion of the Web and, therefore has increased the volume of information available as XML documents. These latter are described through semantic-less tags (eg, Content, Section, Paragraph). Despite the real and urgent need for a semantically exploitation of these XML documents, there are too few research issues tackling the semantic aspects of XML documents. Therefore, the development of automated tools able to determine semantic structures for XML documents is becoming a necessity and a challenge. In this paper, we propose a novel automated approach to create semantic structures for XML documents in order to perform semantic OLAP analyses. The software prototype developed to support this approach is also described.