Manutenção de links na web semântica

Links between data elements described via the RDF model are at the heart of the Semantic Web. The growing number of structured data published in RDF repositories confirms the potential of the global data space proposed by the Semantic Web view. Usually, links are build using (semi)-automatic instance matching algorithms based on pre-defined heuristics and similarity calculation methods. Currently, there is a huge mass of interconnected RDF data that requires methods and tools to deal with their consistency. Although the implementation of change operations in RDF datasets is essential to guarantee the evolution of structured data, such operations can influence established links, which makes the consistency of data connections over time difficult. State-of-the-art studies have studied the broken link problem by devising techniques, which include: dataset versioning; notification of changes to dataset maintainers and backlinks storage. It lacks investigations dealing with the maintenance of links established among RDF datasets. This Ms.c. dissertation investigates new methods and algorithms capable of correcting RDF links between structured data (organized according to ontologies) based on the evolution of the dataset properties. This involves designing techniques and formalizing RDF change operations from one version to another concerning links between instances of ontology concepts. Our work contributes with the following aspects: (1) a survey discussing an extensive study over strategies to debug, discover and fix links inside and outside LOD datasets; (2) a tool to analyze correlations between changes in RDF triples and changes in links (connections to distinct datasets); (3) a technique that automatically discovers semantically broken links based on the evolution of RDF datasets; (4) a (semi)-automatic method that, supported by change operations on RDF datasets, indicates maintenance actions on links, turning inconsistent links healthy again. Our proposal was experimentally evaluated relying on real-world and well-known datasets from the LOD cloud, such as DBpedia, Geonames, Wikidata and Agrovoc. Obtained results reveal that the maintenance of established links can be updated according to RDF change operations affecting resources (AU)