Application datasets are shared or published in a variety of technical formats, structured or semi-structured. Visualization tools can help users explore and understand those data, but are typically designed for one data format, relational, RDF, XML, or even a domain-specific vocabulary such as FOAF for RDF. We present a novel framework that, given as input a dataset of one among many different formats, automatically derives an informative graph structure in the spirit of Property Graphs (PGs), from which an interactive visual interface is created. Our framework leverages prior work on integrating heterogeneous data and abstracting (or summarizing) it. The novelty here lies in interpreting each graph node through this abstraction process, and combining it with an interactive visual representation.