Data Warehouses and the Semantic Web - Data Warehouse Systems: Design and Implementation - page 563

real-world case study, how these operations can be implemented in SPARQL.
Finally, we applied QB4OLAP to the Northwind data cube and query the
resulting RDF cube using SPARQL.
14.7 Bibliographic Notes
There are many topics explaining the basics of the semantic web, for example,
[ 82 ]. A topic entirely devoted to SPARQL is [ 41 ]. At the time of writing
this topic there is no much work on the topic of applying OLAP directly
over RDF data. Section 14.3 of this chapter is based on research work by
Etcheverry and Vaisman on QB4OLAP [ 52 , 53 ]. Kampgen and Harth [ 99 ]
also propose to apply OLAP operations on top of QB, although this approach
does not solve the limitations discussed in this chapter regarding the absence
of dimension structure in QB. In a sequel [ 100 ], the same authors proposed
to load statistical linked data into an RDF triple store and to answer OLAP
queries using SPARQL. For this, they implement an OLAP to SPARQL
engine which translates OLAP queries into SPARQL.
We also mentioned that another research approach studies how to extract
multidimensional data from the semantic web, and then analyze these data
using traditional OLAP techniques. The methods to do this are based on
ontologies, which allow us to extract data in a semiautomatic fashion. The
idea is to use ontologies to identify facts and dimensions that can populate a
data cube. We briefly mention next some of this work.
Niinimaki and Niemi [ 145 ] use ontology mapping to convert data sources
to RDF and then query this RDF data with SPARQL to populate the
OLAP schema. The ETL process is guided by the ontology. In addition,
the authors create an OLAP ontology, somehow similar to the vocabularies
discussedinthischapter.OntologiesareexpressedinRDFandOWL.
Along the same lines, Romero and Abell´o[ 180 ] address the design of the
data warehouse starting from an OWL ontology that describes the data
sources. They identify the dimensions that characterize a central concept
under analysis (the fact concept) by looking for concepts connected to it
through one-to-many relationships. The same idea is used for discovering the
different levels of the dimension hierarchies, starting from the concept that
represents the base level. The output of the method is a star or snowflake
schema that guarantees the summarizability of the data, suitable to be
instantiated in a traditional multidimensional database. Finally, Nebot and
Berlanga [ 142 ] proposed a semiautomatic method for extracting semantic
data on demand into a multidimensional database. In this way, data could
be analyzed using traditional OLAP techniques. Here, the authors assume
that data are represented as an OWL ontology. A portion of this ontology
contains the application and domain ontology axioms, while the other part
contains the actual instance store. A multidimensional schema must first be