Fouilla: a Topical Metadata Improvement for Knowledge Bases derived from Wikipedia

Regarding large knowledge bases, such as Wikipedia, Dbpedia, Freebase, etc., identifying resources relative to a given topic can be a rough task. In such knowledge bases, this kind of relations must appears explicitely in the triples in order to be understandable for machine exploitation. This task is even more complicated if there is a need of ponderation within the topic, which means identify for a given topic the list of articles, sorted importance. At this aim, we proposed a method based on the navigational metadata present in Wikipedia, to compute for a set of automatically identified topics, a list of sorted articles, relevant to the topic in one way or another. We called that method Fouilla.

Based on that method, we levaraged the bijective link between Wikipedia and some derived knowledge bases to provide transversal datasets, in order to enable the utilization of the topical metadata with any Knowledge Base derived from Wikipedia.

To showcase the genericity and the suitability of Fouilla, we selected 3 well-kwown Wikipedia-derived knowledge bases to export the topical ranking computed above in RDF format files, designed to be used simultaenously with the original Knowledge bases. We selected at this aim the following Knowledge bases: Dbpedia, Freebase and Wikidata. This selection looks representative because all of these Knowledge bases differ in their content and utility, but have in common the bilateral link between their entities and Wikipedia articles, which make them suitable for Fouilla.