Multidimensional modeling driven from a domain language

Abstract

The multidimensional model is based on the concepts of facts (business phenomena to be analyzed), dimensions (coordinates for analyzing a fact), hierarchies (descriptions of each dimension at progressively coarser levels of aggregation), and measures (numerical attributes that quantify a fact), and it is commonly adopted for representing data to support the decision-making process. Though multidimensional modeling has been widely investigated, requirements elicitation is still an open issue mainly due to the poor knowledge end-users have of the multidimensional model on the one hand, to the lack of a domain language shared with designers on the other. In the direction of bridging this gap, this paper proposes an approach to obtain a multidimensional schema from the language of the domain captured through a Language Extended Lexicon (LEL). LELs have been introduced as structured glossaries to describe the language used in the application domain, aimed at facilitating requirements elicitation in software engineering. Methods: Our approach consists of two steps. In the first one, end-users apply a set of derivation rules to the LEL in order to obtain draft multidimensional schemata. The second step relies on Multidimensional Modeling Driven From a Domain Language the interaction of end-users and designers to review and edit these draft multidimensional schemata so as to obtain the final ones. Results: The approach is validated via an experiment made on a case study, showing that end-users who apply our rules tend to produce multidimensional schemata that are more correct than those produced by end-users who work freely. Conclusion: Our rules provide a structured context where subjectivity has a smaller impact than in the case of designing with no guidelines, thus effectively supporting the collaboration between end-users and designers.