Towards a linked-data based visualization wizard
Ghislain Auguste Atemezing, Raphaël Troncy · 2014
Abstract. Datasets published in the LOD cloud are recommended to follow some best practice in order to be 4-5 stars Linked Data compli-ant. They can often be consumed and accessed by different means such as API access, bulk download or as linked data fragments, but most of the time, a SPARQL endpoint is also provided. While the LOD cloud keeps growing, having a quick glimpse of those datasets is getting harder and there is a need to develop new methods enabling to detect auto-matically what an arbitrary dataset is about and to recommend visu-alizations for data samples. We consider that “a visualization is worth a million triples”, and in this paper, we propose a novel approach that mines the content of datasets and automatically generates visualizations. Our approach is directly based on the usage of SPARQL queries that will detect the important categories of a dataset and that will specifically con-sider the properties used by the objects which have been interlinked via owl:sameAs links. We then propose to associate type of visualization for those categories. We have implemented this approach into a so-called