This research forms part of Petrelli’s enquiry into the human interaction with extra-large data sets and the design of visual mechanisms to support the user in making meaning out of millions of entities. For this research, the challenge was to find a strategy for the automatic visualisation of any dataset that could come from any source and can be dynamically linked to any other datasets in a web of data known as Linked Data. Because of the automatic aggregation, naturalistic dimensions such as a timeline or a geographical map cannot be used as the type of data is unknown. The research goal was then to define a set of abstract visual features to automatically generate an easy-to-use and pleasing interactive visualisation of any unknown dataset. Given the huge size of the dataset, the research sought to design an interactive visual summary that enabled the user to understand, at a glance, an unknown dataset and invite the interactive exploration of millions of entries via visual query. To address the visualisation complexity and to support engagement and exploration, principles of graphic design were included in the algorithm to prove the statement that a ‘pleasing’ user interface is more engaging. This proposition was a clear departure from the visualisations developed by the Semantic Web community and proved superior to four other systems in a comparative evaluation. By describing in detail the process followed in implementing the design rationale, the paper enabled other researchers in the Semantic Web community to adopt clear and effective principles for the visualisation of and user interaction with linked-data addressing open issues of usability, as noted in reviews of this research. This paper was selected by the journal editorial board of the Semantic Web Journal to be presented at the International Semantic Web Conference, October 2016, Kobe, Japan.