The RDF format offers powerful possibilities for machines, such as reasoning or federated queries over interlinked datasets. However, presenting RDF data to humans is very challenging: its very structure defeats traditionnal approaches, as it separates information into small pieces, making it difficult for users to make sense of it. My PhD work proposes an approach that presents RDF data in a context, to make them understandable by humans. We first describe S-Paths, a system to support set-based exploration of a dataset's content. We show that it works well on simple models, but that its efficiency is limited by performance issues on very abstract models. Then we lay the basis for a second project, whose aim is to take one more step back and put these sets of entities in a broader context, to give a structural overview of Linked Datasets.