Bridging Property Graphs and RDF for IoT Information Management Abdullah Abbas 1 and Gilles Privat 2 Orange Labs, Grenoble, France 1 abdullah.abbas@orange.com 2 gilles.privat@orange.com Abstract. “Property graphs” have come to be widely used as the common me- ta-model of most graph databases. The anterior prevalence of the RDF meta- model, its rich suite of tools and its openness to the linked data cloud, make it essential to ensure the consistency and potential conversion between these two models. We propose to give property graphs a formal semantic grounding based on RDF/RDFS/OWL, with blank nodes reification, geared to JSON-LD seriali- zation. On top of it, we also define a set of core cross-domain ontology classes for the IoT, based on this meta-model, intended to be used for IoT platforms alongside the proposed meta-model. Keywords: IoT, Web of things, RDF, Property graphs, Meta-model, Ontology, Reification. 1 Introduction Internet of Things (IoT) platforms, overly numerous as they are [1], have failed to converge on any semblance of a common information model. They mostly rely on low-level ad hoc data models, implicitly object-based and hierarchical, with weak un- formalized semantics, falling short of requirements for inter-operation with other platforms and potential sharing of data. Beyond legacy solutions, an evolution that points in the right direction is towards the increasing adoption, though more so indus- try than academia, of the “property graph” meta-model, the common denominator of most graph databases that are already widely used in the industry. As an evolution of rigid relational database schemata, property graphs [2, 3] provide an obvious ad- vantage in flexibility. Yet they are also highly relevant as an evolution of object- oriented information models that are implicitly based on an underlying arborescent data structure (directed rooted tree), whereas an unrestricted graph is vastly more versatile and expressive. Such a full-fledged graph-based model is the best candidate for a shared information model in the IoT and Cyber-Physical Systems. For this, it should capture not only the semantics of the individual descriptions of IoT and CPS components, but first and foremost, their structure and potentially their behavior, as we explain in the following sections. 77