fabiovalse

fabiovalse

fabiovalse

This experiment shows the results of some classifier algorithm. Each bar corresponds to a different algorithm. In particular this bar chart tries to visualise and make comparable the four statistical numbers of binary classification: True Positive (TP), True Negative (TN), False Positive (FP) and False Negative (FN). Each of these components are represented in the diagram using bars with different color and height:

• TP are green and high;
• TN are red and high;
• FP are green and small;
• FN are red and small.

Moreover, it shows the accuracy of the algorithms as the union of the high bars (TP + TN). The first bar in the diagram represents the ideal algorithm in which there are only TP and TN and no FP and FN.

fabiovalse

This experiment displays the graph of the classes of DBpedia. Each node is a class (e.g., Person, Organisation, Event, Place, Species) while each link represents the amount of triples between the entities of the classes it connects.

Given a node, for instance the class Person, it is important to specify that it represents all the entities having as class (rdf:type) Person without any of its subclasses. That is, each instance is represented only with the most specific class it is classified in the DBpedia hierarchy (ontology).

fabiovalse

This experiment displays the graph of the classes of DBpedia and improves the results obtained in the previous node link diagram. Each node is a class (e.g., Person, Organisation, Event, Place, Species) while each link represents the amount of triples between the entities of the classes it connects.

Given a node, for instance the class Person, it is important to specify that it represents all the entities having as class (rdf:type) Person without any of its subclasses. That is, each instance is represented only with the most specific class it is classified in the DBpedia hierarchy (ontology).

fabiovalse

This experiment extends the previous radial tree by adding hierarchical bundling. The main link flows can be easily identified in the diagram looking at the darker lines denoting strong links overlaps.

fabiovalse

This variant of the hierarchical bundling has links grouped in pipes. The idea was inspired by the Strahler number.

fabiovalse

This experiment is an evolution of the previous one. The links are shown as outcoming links (red) and incoming links (blue). Moreover, the size of the nodes describes the number of entities for a certain DBpedia Class.

fabiovalse

This experiment shows the mathematical and astronomical lexicon of the Clavius project. The sunburnst shows the classes of the project Ontology. The root (#Lexical_Entry) and the other classes of the ontology are depicted in lightblue. The leaves that represents the entities are depicted in blue.

fabiovalse

This experiment shows a little part of DBpedia: the Event sub-hierarchy. It represents it as a map in which the regions are the ontological classes in the sub-hierarchy. Moreover, in order to create the regions, a GOSPER curve has been used for the cell displacement. Each cell has an hexagonal shape and it represents a DBpedia entity. Hence, the area of each region shows the size of a class in the overall Event hierarchy. The different region lines (solid, dashed and with different thickness) give the class depth in the hierarchy.

The map shows 45377 cells (entities) and 24 regions (ontological classes).

Future works will include:

• Information about the cells;

• Region labeling;