Scalable methods to analyze Semantic Web data

Semantic Web data is currently being heavily used as a data representation format in scientific communities, social networks, business companies, news portals and other domains. The irruption and availability of Semantic Web data is demanding new methods and tools to efficiently analyze such data and take advantage of the underlying semantics. Although there exist some applications that make use of Semantic Web data, advanced analytical tools are still lacking, preventing the user from exploiting the attached semantics. The main objective of this dissertation is to provide a formal framework that enables the multidimensional analysis of Semantic Web data in an scalable and efficient manner. The success of multidimensional analysis techniques applied to large volumes of structured data in the context of business intelligence, especially for data warehousing and OLAP applications, has prompted us to investigate the application of such techniques to Semantic Web data, whose nature is semi-structured and contain implicit knowledge. Multidimensionality is based on the fact/dimension dichotomy. Data are modeled in terms of facts, which are the analytical metrics, and dimensions, which are the different analysis perspectives, and are usually hierarchically organized. We believe that the construction of a multidimensional view of Semantic Web data driven by the semantics encoded in the data themselves and the user’s requirements, empowers and enriches the analysis process in a unique manner, as it brings about new analytical capabilities not possible before. Aggregations and display operations typical of multidimensional analysis tools, such as changing the granularity level of the displayed data, or adding a new analysis perspective to the data, will be performed based on the semantic relations encoded in the data. This is possible thanks to the mapping of the data to a conceptual multidimensional space. We base our research on the hypothesis that Semantic Web data is an emerging knowledge resource worth exploiting, and that the knowledge encoded in Semantic Web data can be leveraged to perform an efficient, scalable and full-fledged multidimensional analysis. Scalability is achieved by two means. On one hand, we provide