Jamel Feki— Ines Ben-Messaoud

University of Sfax, Faculty of Economics and Management, Computer Department,
Multimedia InfoRmAtion Computing Laboratory (MIRACL)

 Abstract
This paper presents a natural language oriented software tool for the reuse of multidimensional pattern (MP). An MP is a star schema describing a generic solution of typical decisional requirements in a given domain. In order to facilitate the reuse of an MP, we hide its structure by generating two types of natural language queries: Dimension queries and Hierarchy queries. Dimension queries serve to select relevant dimensions in the MP and, hierarchy queries, i.e., queries expressed on dimensional attributes help to decide which analytical levels interest the analysts (i.e., decision-maker). To assist the analysts focusing on the most significant queries, we assign each query a calculated weight representing its importance in the analyses. We present a toolset, called NL-PI that implements our Natural Language Pattern Instantiation method and, we illustrate its functionalities with an example in the commercial domain.     

Keywords: Multidimensional pattern, decision support system, data mart construction, natural language instantiation, design reuse.