Uma abordagem para a geração semiautomática de mapeamentos R2R baseado em um catálogo de padrões

Abstract

The web of linked data has grown considerably in recent years and covers a wide range of different domains today (BIZER; JENTZSCH; CYGANIAK, 2011). Linked data sources use different vocabularies to represent data about a specific type of object. For example, DBpedia 3 and Music ontology 4 use their proprietary vocabularies to represent data About musical artists. Translating data from these bound data sources into the vocabulary that is expected by a linked data application requires a large number of mappings and may require many structural transformations as well as complex transformations in the property value. Several tools emerge to map ontologies such as the SPARQL 1.1 language, LDIF framework and the Mosto tool. We choose to use in our study the R2R language, which was pointed out in (BIZER et al., 2012) as a good option to map ontologies, as it stands out in terms of expressiveness and performance. The R2R mapping language is a language based on the SPARQL language that allows you to transform data from a source vocabulary into a user-defined target vocabulary. However, defining mappings using this language is complex and subject to several types of errors, such as writing errors or even semantic errors, requiring expect user most to define the mappings. In this scenario, we propose an approach, using mapping patterns to automatically generate R2R mappings from a AMs. The approach is divided into two steps: (1) the manual specification of a set of AMs between the vocabulary of a source ontology and the vocabulary of a target ontology of the user’s choice; and (2) the automatic generation of the R2R mappings based on the result of the first step. Finally, we present the R2R By Assertions tool to help the user in the process of generating R2R mapping.