Abstract
International audience; As the amount of textual information grows explosively in various kinds of business systems, it becomes more and more essential to analyze both structured data and unstructured textual data simultaneously. However information contained in non structured data (documents and so on) is only partially used in business intelligence (BI). Indeed On-Line Analytical Processing (OLAP) cubes which are the main support of BI analysis in decision support systems have focused on structured data. This is the reason why OLAP is being extended to unstructured textual data. In this paper we introduce the innovative “Diamond” multidimensional model that will serve as a basis for semantic OLAP on XML documents and then we describe the meta modeling, generation and implementation of a the Diamond multidimensional model.