Semantic Metadata for Supporting Exploratory OLAP

Abstract

On-Line Analytical Processing (OLAP) is an approach widely used for data analysis. OLAP is based on the multidimensional (MD) data model where factual data are related to their analytical perspectives called dimensions and together they form an n-dimensional data space referred to as data cube. MD data are typically stored in a data warehouse, which integrates data from in-house data sources, and then analyzed by means of OLAP operations, e.g., sales data can be (dis)aggregated along the location dimension. As OLAP proved to be quite intuitive, it became broadly accepted by non-technical and business users. However, as users still encountered difficulties in their analysis, different approaches focused on providing user assistance. These approaches collect situational metadata about users and their actions and provide suggestions and recommendations that can help users' analysis. However, although extensively exploited and evidently needed, little attention is paid to metadata in this context. Furthermore, new emerging tendencies call for expanding the use of OLAP to consider external data sources and heterogeneous settings. This leads to the Exploratory OLAP approach that especially argues for the use of Semantic Web (SW) technologies to facilitate the description and integration of external sources. With data becoming publicly available on the (Semantic) Web, the number and diversity of non-technical users are also significantly increasing. Thus, the metadata to support their analysis become even more relevant. This PhD thesis focuses on metadata for supporting Exploratory OLAP. The study explores the kinds of metadata artifacts used for the user assistance purposes and how they are exploited to provide assistance. Based on these findings, the study then aims at providing theoretical and practical means such as models, algorithms, and tools to address the gaps and challenges identified. First, based on a survey of existing user assistance approaches related to OLAP, the thesis proposes the analytical metadata (AM) framework. The framework includes the definition of the assistance process, the AM artifacts that are classified in a taxonomy, and the artifacts organization and related types of processing to support the user assistance. Second, the thesis proposes a semantic metamodel for AM. Hence, Resource Description Framework (RDF) is used to represent the AM artifacts in a flexible and re-usable manner, while the metamodeling abstraction level is used to overcome the heterogeneity of (meta)data models in the Exploratory OLAP context. Third, focusing on the schema as a fundamental metadata artifact for enabling OLAP, the thesis addresses some important challenges on constructing an MD schema on the SW using RDF. It provides the algorithms, method, and tool to construct an MD schema over statistical linked open data sets. Especially, the focus is on enabling that even non-technical users can perform this task. Lastly, the thesis deals with queries as the second most relevant artifact for user assistance. In the spirit of Exploratory OLAP, the thesis proposes an RDF-based model for OLAP queries created by instantiating the previously proposed metamodel. This model supports the sharing and reuse of queries across the SW and facilitates the metadata preparation for the assistance exploitation purposes. Finally, the results of this thesis provide metadata foundations for supporting Exploratory OLAP and advocate for greater attention to the modeling and use of semantics related to metadata.