An identification ontology for entity matching

Abstract

In this context we present the Identification Ontology, as an application ontology for a knowledge-based solution to the entity matching problem in the context of the Semantic Web. The Identification Ontology has a threefold role: (1) represent a selection of attributes that are relevant for identification (or entity matching) of a set of entity types; (2) supporting the definition of a set of contextual ontological mappings to ease the problem of semantic heterogeneity affecting entity matching in the Semantic Web; and (3) represent meta-properties of the considered features to highlight their roles in the definition of a knowledge-based entity matching solution. The Identification Ontology taxonomy is defined refining and extending the Okkam Conceptual Model, as a top level ontology modeling the identity and reference domain. Furthermore, it defines also a set of top-level entity types and relative features relying on a methodology that combines results from cognitive studies and a survey of existing vocabularies available through Linked Open Vocabulary initiative. The Identification Ontology is currently used as part of the Okkam Entity Name System matching framework, which was successfully tested in entity matching experiments and used in large-scale (industrial) linkage tasks to enable data integration for applications dealing with tax assessment and credit risk analysis.