The true value of Linked Data becomes apparent when datasets are analyzed and understood already at the basic level of data types, constraints, value patterns etc. Such data pro ling is especially challenging for Rdf data, the underlying data model on the Web of Data. In particular, graph analysis can be used to gain more insight into the data, induce schemas, or build indices. We present ProLod++, a tool for various pro ling and mining tasks and in particular its recent extension GraphLod, which o ers Rdf graph analysis features. ProLod++ features many interactive pro ling results speci c for open data, such as schema discovery for user-generated attributes, association rule discovery to uncover synonymous predicates, and key discovery along ontology hierarchies. GraphLod enhances it with subgraph pattern mining, node degree distribution, component visualization and analysis, and more.
generated attributes, association rule discovery to uncover synonymous
predicates, and key discovery along ontology hierarchies [
{ Interactive graph structure exploration in a faceted fashion. 2
The features of ProLod++ can be categorized into pro ling and mining tasks, as illustrated in Table 1.
Basic Analysis. Imported data is clustered by hierarchical topic clustering if no
underlying schema is available, otherwise it is grouped based on the underlying
taxonomic hierarchy. The pro ling and mining tasks are executed on import and
results are stored in a relational database. These include statistics on frequencies
and distributions of distinct subjects, properties, and objects. Pattern analysis
provides the user with statistics on data types and value pattern distributions
of particular properties. ProLod++ discovers positive and negative association
rules, e.g., to discover synonymous properties or inverse properties. To cope
with the sparsity of property values on the Web of Data when discovering key
candidates, ProLod++ calculates the keyness measure for each property along
the ontology class hierarchy. These features were already demonstrated in [
ProLod++ allows faceted browsing through the graph patterns. Patterns are grouped when isomorphic, rst based on their underlying structure and then based on the class membership (color). This allows for nding not only common, re-occurring patterns but also patterns that are dominant for certain class-combinations. E.g., astronomers in DBpedia are often to be found in star patterns, surrounded by their discovered astronomical objects.
Based on the graph features provided by ProLod++ and its underlying GraphLod library, an overall model for Linked Datasets can be given: We observe that most of the Linked Datasets consist of a number of small satellite graphs and a giant component that contains more than 80% of the nodes and thus resemble scale-free networks as they occur in social networks.
When jointly pro ling multiple datasets, ProLod++ highlights the connectivity of connected components across them based on inter-dataset links. This, for instance, identi es the potential of dataset integration. 3
ProLod++ is a web-based tool to be either distributed for local execution or hosted as a service at http://prolod.org. Some of the described features are still under development, but at the time of submission ProLod++ is already a useful tool to explore Rdf datasets and their graph structure. During the demo, users can bring along their own Rdf dataset, import it into ProLod++ and begin the analysis. A number of several interesting datasets from various domains have been already imported, including DBpedia, Diseasome, and LinkedMDB.
After the initial analysis phase, users can select datasets and clusters in a tree model and browse the pro ling results across several tabs. The graph feature analysis shows graph statistics, such as number of nodes and edges, and the diameter for the connected and strongly connected components. A node degree distribution chart is displayed to analyze the underlying graph model. Besides statistical information, ProLod++ allows faceted browsing through the graph patterns, from general patterns to class-colored patterns down to concrete pattern examples. The class distribution is visualized at each facet level.