Task-Oriented Uncertainty Evaluation for Linked Data Based on Graph Interlinks

Description

For data sources to ensure providing reliable linked data, they need to indicate information about the (un)certainty of their data based on the views of their consumers. In Addition, uncertainty information in terms of Semantic Web has also to be encoded into a readable, publishable, and exchangeable format to increase the interoperability of systems. This paper introduces a novel approach to evaluate the uncertainty of data in an RDF dataset based on its links with other datasets. We propose to evaluate uncertainty for sets of statements related to user-selected resources by exploiting their similarity interlinks with external resources. Our data-driven approach translates each interlink into a set of links referring to the position of a target dataset from a reference dataset, based on both object and predicate similarities. We show how our approach can be implemented and present an evaluation with real-world datasets. Finally, we discuss updating the publishable uncertainty values.

Abstract

International audience

Additional details