An Ontology-Based Method for Duplicate Detection in Web Data Tables

International audienceWe present, in this paper, a duplicate detection method in semantically annotated Web data tables, driven by a domain Termino- Ontological Resource (TOR). Our method relies on the fuzzy semantic annotations automatically associated with the Web data tables. A fuzzy semantic annotation is automatically associated with each row of a Web data table. It corresponds to the instantiation of a composed concept of the domain TOR, which represents the semantic n-ary relationship that exists between the columns of the Web data table. A fuzzy semantic annotation contains fuzzy values expressed as fuzzy sets. We propose an automatic duplicate detection method which consists in detecting the pairs of duplicate fuzzy semantic annotations and relies on (i) knowledge declared in the domain TOR and on (ii) similarity measures between fuzzy sets. Two new similarity measures are defined to compare both, the symbolic fuzzy values and the numerical fuzzy values. Our method has been tested on a real application in the domain of chemical risk in food