MLED_BI: A Novel Business Intelligence Design Approach to Support Multilingualism

With emerging markets and expanding international cooperation, there is a requirement to support Business Intelligence (BI) applications in multiple languages, a process which we refer to as Multilingualism (ML). ML in BI is understood in this research as the ability to store descriptive content (such as descriptions of attributes in BI reports) in more than one language at Data Warehousing (DWH) level and to use this information at presentation level to provide reports, queries or dashboards in more than one language.
Design strategies for data warehouses are typically based on the assumption of a single language environment. The motivations for this research are the design and performance challenges encountered when implementing ML in a BI data warehouse environment. These include design issues, slow response times, delays in updating reports and changing languages between reports, the complexity of amending existing reports and the performance overhead. The literature review identified that the underlying cause of these problems is that existing approaches used to enable ML in BI are primarily ad-hoc workarounds which introduce dependency between elements and lead to excessive redundancy. From the literature review, it was concluded that a satisfactory solution to the challenge of ML in BI requires a design approach based on data independence the concept of immunity from changes and that such a solution does not currently exist.
This thesis presents MLED_BI (Multilingual Enabled Design for Business Intelligence). MLED_BI is a novel design approach which supports data independence and immunity from changes in the design of ML data warehouses and BI systems. MLED_BI extends existing data warehouse design approaches by revising the role of the star schema and introducing a ML design layer to support the separation of language elements. This also facilitates ML at presentation level by enabling the use of a ML content management system. Compared to existing workarounds for ML, the MLED_BI design approach has a theoretical underpinning which allows languages to be added, amended and deleted without requiring a redesign of the star schema; provides support for the manipulation of ML content; improves performance and streamlines data warehouse operations such as ETL (Extract, Transform, Load). Minor contributions include the development of a novel BI framework to address the limitations of existing BI frameworks and the development of a tool to evaluate changes to BI reporting solutions.
The MLED_BI design approach was developed based on the literature review and a mixed methods approach was used for validation. Technical elements were validated experimentally using performance metrics while end user acceptance was validated qualitatively with end users and technical users from a number of countries, reflecting the ML basis of the research. MLED_BI requires more resources at design and initial implementation stage than existing ML workarounds but this is outweighed by improved performance and by the much greater flexibility in ML made possible by the data independence approach of MLED_BI. The MLED_BI design approach enhances existing BI design approaches for use in ML environments.