Overall Objectives
Research Program
Application Domains
Software and Platforms
Partnerships and Cooperations
Bibliography
 PDF e-Pub

Section: New Results

The web of data uses semantic web technologies to publish data on the web in such a way that they can be interpreted and connected together. It is thus critical to be able to establish links between these data, both for the web of data and for the semantic web that it contributes to feed. We consider this problem from different perspectives.

Participants : Tatiana Lesnikova [Correspondent] , Jérôme David, Jérôme Euzenat.

Data interlinking is a difficult task in a cross-lingual environment like the Web. Even systems based on graph structure, ultimately rely on anchors based on language fragments. If languages are different, fragments have to be compared by more sophisticated techniques. In that context, we are developing an approach which represents RDF entities as (virtual) text documents and compare them using different strategies [9] , [10] . We investigate two directions: (1) a translation-based approach where the virtual documents are automatically translated; (2) a language-independent approach where important terms found in documents are mapped to a terminological resource like Wordnet to compute document similarity.

This work is part of the PhD of Tatiana Lesnikova developed in the Lindicle project (see § 8.1.2 ).

Participants : Zhengjie Fan [Correspondent] , Jérôme Euzenat.

In the context of the Datalift project, we are further developing the data interlinking module. We have developed an algorithm able to determine potential attribute correspondences of two classes depending on their features. For that purpose, we use $k$-means or $k$-medoids clustering. These correspondences are then used to construct a Silk script which generates an initial link set. Some of the links are presented to the user who assesses their validity. We then use an improvement of the disjunctive version space supervised learning method to learn a better script from the assessed links. Such a technique can be iterated until satisfactory links are found.

Key and pseudo-key detection for web data set interlinking

Participants : Jérôme David [Correspondent] , Manuel Atencia Arcas, Anthony Delaby, Jérôme Euzenat.

Keys are sets of properties which uniquely identify individuals (instances of a class). We have refined the notion of database keys in a way which is more adapted to the context of description logics and the openness of the semantic web. We have also refined the weaker notion of a linkkey introduced in [12] . Then we have shown how such keys, together with ontology alignments, and linkkeys may be used for deducing equality statements (links) between individuals across data sources in the web of data.

However, ontologies do not necessarily come with key descriptions, and never with linkkey assertions (which would hold across ontologies). But, these can be extracted from data by assuming that keys holding for specific data sets, may hold universally. We have extended these classical key extraction techniques for extracting linkkeys.

This work is developed partly in the Lindicle and Datalift projects. A proof of concept implementation is available at http://rdfpkeys.inrialpes.fr/ .