A natural language-based approach for a semi-automatic data mart design and ETL generation

Data warehousing projects still face challenges in the various phases of the development life cycle. In particular, the success of the design phase, the focus of this paper, is hindered by the cross-disciplinary competences it requires. This paper presents a natural language-based method for the design of data mart schemas. Compared to existing approaches, our method has three main advantages: first, it facilitates requirements specification through a template covering all concepts of the decision-making process while providing for the acquisition of analytical requirements written in a structured natural language format. Second, it supports requirement validation with respect to a data source used in the ETL process. Third, it provides for a semi-automatic generation of conceptual data mart schemas that are directly mapped onto the data source; this mapping assists the definition of ETL procedures. The performance of the proposed method is illustrated through a software prototype used in an empirical comparison of our method with top-down methods.