Exploiting Resource Description Framework (RDF) data for Online Analytical Processing (OLAP), especially Linked Open Data (LOD), could allow analysts to obtain interesting insights. To conduct OLAP analysis over RDF data, analysts should know the speci c semantics, structure, and querying mechanisms of such data. Furthermore, these data should ideally adhere to a multidimensional structure to be accessible to OLAP. In this demo paper, we present an OLAP endpoint that allows casual analysts to perform self-service OLAP analysis over RDF datasets. Speci cally, analysts can instantiate semantic web analysis graphs, which are prede ned models of the analysis processes. Semantic web analysis graphs are built on top of multidimensional structures that can be superimposed over arbitrary RDF datasets.
Online Analytical Processing (OLAP) systems support business analysts in their
decision-making processes by allowing them to view the data at di erent
granularities. In order to be accessible to OLAP, these data are organized in a
multidimensional (MD) structure. An MD model consists of facts which are the
subjects of the analysis and quanti ed by measures, and hierarchically-organized
dimensions allowing for measure aggregation. Traditionally, OLAP systems have
targeted enterprise-internal data. Yet, external RDF data, such as DBpedia
and Wikidata, are an important source of knowledge that is still largely
unexploited for OLAP analysis. External RDF data, however, do not correspond
to a structure easily accessible to OLAP systems. The superimposition of
analytical schemas is, therefore, a common strategy to render these data accessible
to OLAP [
In this demo paper1, we present a self-service OLAP endpoint [
Multidimensional schemas (MDS) can be superimposed over arbitrary RDF datasets to express MD analytical possibilities. Superimposition means that the data do not originally follow an MD structure, but this structure is rather imposed on top of data that are left unchanged. A mapping, however, relates MDS to the data in a GLAV (global-as-view) fashion using SPARQL queries. An OWL ontology is established to express the MDS metamodel. This metamodel can be instantiated to express MD schemas for particular RDF sources. The upper part of Fig. 1 shows an example instantiation of the MDS metamodel for the news example. Mention is the fact. Source, Category, Entity, and Time are dimensions with some having hierarchies. ConfScore and NumOfMentions are measures. Entity dimension illustrates specialization as MentionedEntity is specialized to Person and Organizationn, which is further specialized to Company. Company rolls up to Industry and has number of employees NumOfEmps as an attribute. 1 A video demonstrating the developed prototype can be found at https://youtu.
be/ymhkqla8J1I 2 http://ontotext.com/semantic-solutions/dynamic-semantic-publishingplatform/linked-data-integration-for-global-publishers/ 3 http://factforge.net/sparql
An OLAP Endpoint for RDF Data Analysis Using Analysis Graphs
We extend Analysis Graphs, proposed for relational data warehousing [
Descriptor Dimension
Level Fact subLevelOf hasMeasure hasLevel rol sUpTo Measure
Category CategoryOfNews NumOfMentions
ConfScore
Month
Source SourceWebSite Mention Figure 2 illustrates the analysis process using analysis graphs. The analyst rst selects an analysis situation to start from in the analysis graph. Then, the analyst can view the analysis situation speci cation and unbound parameters and bind them supported by auto-completion. Once the form is submitted, a SPARQL query which corresponds to the analysis situation is generated and sent to the data source's endpoint. The results are then retrieved and can be viewed in tabular form or as charts. Afterwards, the analyst can select a navigation to one of the subsequent analysis situations. Once a navigation is performed, the target analysis situation is activated with the operations corresponding to the performed navigation. The graphical user interface facilitates the process and guides the user. Figure 3 sketches the system architecture. The OLAP Endpoint is implemented as Java Server Pages (JSP). D3 force layout is used to visualize the analysis graph, while Google charts are used to visualize the results. The endpoint is controlled via the Controller. The overall analysis process is coordinated and performed by the Analysis Graphs Execution Engine. Parsing and reasoning over MDS and SWAG ontologies and their instances is performed by OWL Handler, which is implemented using Jena libraries. SPARQL Query Creator is implemented using Jena ARQ library, and is responsible for generating
Specification
Start select analysis situation navigate parameters
Analysis Graph query
Results query back to analysis
graph select navigation SPARQL queries that correspond to the MD queries speci ed by bound analysis situations. Connection Manager is responsible for accessing external required les and interacting with the dataset Endpoint.
interacts OLAP Endpoint
SPARQL Query
Creator
Controller
Analysis Graphs Execution
Engine OWL Handler
Connection Manager
System
We plan to enable analysts to manipulate analysis graphs on the y, thus extending or editing the analysis situations and navigations at runtime or navigating without a prede ned schema. Furthermore, we plan to conduct a user study to evaluate the usability of the approach and improve it correspondingly.