Ontology-Based Conceptual Design of ETL Processes for Both Structured and Semi-Structured Data

Abstract

One of the main tasks in the early stages of a Data Warehouse project is the identification of the appropriate transformations and the specification of inter-schema mappings from the data sources to the Data Warehouse. In this paper, we propose an ontology-based approach to facilitate the conceptual design of the back stage of a Data Warehouse. A graph-based representation is used as a conceptual model for the datastores, so that both structured and semi-structured data are supported and handled in a uniform way. The proposed approach is based on the use of Semantic Web technologies to semantically annotate the data sources and the Data Warehouse, so that mappings between them can be inferred, thereby resolving the issue of heterogeneity. Specifically, a suitable application ontology is created and used to annotate the datastores. The language used for describing the ontology is OWL-DL. Based on the provided annotations, a DL reasoner is employed to infer semantic correspondences and conflicts among the datastores and propose a set of conceptual operations for transforming data from the source datastores to the Data Warehouse.

Ontology-Driven Conceptual Design of ETL Processes Using Graph Transformations

“…However, to the best of our knowledge, the approaches presented in [10,18] were the first to explore the application of ontologies to the conceptual design of ETL processes. In these works, the schemata of the source and target data stores are annotated by complex expressions built from concepts of a corresponding domain ontology.…”

“…A first attempt towards this direction has been recently presented [10]. In this paper, we build upon the idea of using an ontology for the conceptual design of ETL processes, and, more specifically, we elaborate on that by proposing a formal way for deriving a conceptual ETL design, based on the well-established graph transformation theory.…”

A Unified View of Data-Intensive Flows in Business Intelligence Systems: A Survey

“…On the one hand, in [88], the authors introduced the design approach based on Semantic Web technologies to represent the DW domain (i.e., source and target data stores), showing that this would further enable automation of the design process by benefiting from the automatic reasoning capabilities of an ontology. [79], on the other hand, assumes that only data sources are captured by means of a domain ontology with associated source mappings.…”

“…For example, the idea of creating a unified view over relevant data sources (i.e., the main principle of the data integration setting), is revisited by some approaches by creating a common domain vocabulary and integrating it with existing data sources (e.g., [88,79]). There, the use of a domain ontology to reconcile the languages of business and IT worlds when building a BI system has been proposed.…”

Ontology driven search of compound IDs

“…In these cases, or even if there is not any conceptual formalization of the domain available we can use integration and reverse engineering techniques for building the needed ontology. For example, (Skoutas and Simitsis, 2007) shows that structured as well as unstructured data stores can be elegantly represented as graphs, and it describes how an ontology for such data stores can be semi-automatically constructed by integrating a domain vocabulary with the data sources' vocabulary. An alternative to the previous approach is presented in (Dánger and Berlanga, 2009).…”