Semantic Web Applications can only be understood if the complex data flows they implement are clearly described. However, application developers have very little support at the moment for documenting such data flows and their rationale, in an appropriately formal and conceptual manner. In this paper, we propose to apply a knowledge engineering approach to the formal description of Semantic Web Ap-plications. Following an ontology building methodology based on the analysis of several existing Semantic Web Applications, we devise an abstract, foundational model of data flow descriptions which fo-cuses on the graph of relationships between “data objects ” within the application, rather than the tasks and processes being implemented. The result is Datanode, a conceptual framework designed to describe systems by expressing relations between data nodes (objects), imple-mented as an ontology of the possible relationships between such data nodes. We evaluate Datanode by applying it to the description of eight recent Semantic Web Applications.

Abstract. Data is at the center of current developments on the World Wide Web. In the Semantic Web, different kinds of data artifacts (datasets, catalogues, provenance metadata, etc.) are published, exchanged, described and queried every day. Data hubs are also emerging in the context of the web of Linked Data, as a way to manage this heterogeneity. There are a number of use cases related to data hub management that can only be addressed if we are able to specify the relations between the managed data artifacts in a way that support useful inferences. This includes the understanding of how the features of the data artifacts propagate. This may not be trivial if we consider complex relations, possibly including datasets in different repositories or data flows happening in independent processes and workflows. We propose an abstract, foundational model which focuses on the graph of relationships between generic “data objects ” (which we call datanodes). Following an ontology building methodology based on the analysis of Semantic Web applications, we devise a foundational Ontology Design Pattern by collapsing different types of data objects together, and by remodelling structured relations to simple binary relations. Our pattern represents a datanode related with a datanode, where the relation can be specified in six fundamental ways. We extend this foundational model and propose a conceptual framework designed to express relations between data nodes, implemented as an extendable OWL ontology covering the possible relationships between such data nodes. We show how the Datanode approach can support the management of (possibly distributed and interlinked) catalogues of web data and reasoning over the relationships between items in data catalogues.