Section: New Results

DBpedia in French

Participants : Julien Cojan, Fabien Gandon.

The purpose of the project DBpedia in French is to extract data from Wikipedia in French and publish it under structured format. Wikipedia content is mainly meant to be read by human and is not suited for use in applications. DBpedia publishes the data extracted from Wikipedia articles in RDF W3C standard for the Semantic Web (http://www.w3c.org/RDF/ ) thus readily available for many applications. For instance, DBpedia is used to generate indexes for cultural resources (e.g. HdA-lab project (http://hdalab.iri-research.org/hdalab/ )), it can also be used for mobile applications thanks to the geographic data it contains, or to answer natural language questions, etc.

The original version of DBpedia is focused on the English chapter of Wikipedia. Last versions also contain elements extracted from other chapters, but only when related to a page in English. Articles with no equivalent in English are skipped, leading to a significant number of pages being ignored and so a significant amount of data is lost. For instance, about 49 000 persons and 180 000 places described in the French chapter have no corresponding article in English and are then missing in the English DBpedia. Moreover, the description of the same topic can be different from one chapter to another, reflecting cultural diversity.

DBpedia in French publishes data extracted from the French Wikipedia in complement to the English DBpedia. Data are linked with the different chapters from the internationalization committee thus providing multilingual resources. In its release from October $2^{\text{nd}}$, DBpedia in French contains 130 million triples describing 1.3 million things, among them 260 000 places, 140 000 persons, 64 000 work pieces and 26 000 organizations.

This project is supported by the Semanticpedia collaboration platform (http://semanticpedia.org ) launched November ${19}^{\text{th}}$ 2012 by Aurélie Filipetti, the French Ministry of Culture, Michel Cosnard, CEO of Inria, and Rémi Mathis, CEO of Wikimédia France. Inria currently hosts the project (http://fr.dbpedia.org ) and is the correspondent for the French chapter in DBpedia internationalization committee.