Managing and Compiling Data Dependencies for Semantic Applications Using Databus Client

Abstract

Realizing a data-driven application or workflow, that consumes bulk data files from the Web, poses a multitude of challenges ranging from sustainable dependency management supporting automatic updates, to dealing with compression, serialization format, and data model variety. In this work, we present an approach using the novel Databus Client, which is backed by the DBpedia Databus - a data asset release management platform inspired by paradigms and techniques successfully applied in software release management. The approach shifts effort from the publisher to the client while making data consumption and dependency management easier and more unified as a whole. The client leverages 4 layers (download, compression, format, and mapping) that tackle individual challenges and offers a fully automated way for extracting and compiling data assets from the DBpedia Databus, given one command and a flexible dependency configuration using SPARQL or Databus Collections. The current vertical-sliced implementation supports format conversion within as well as mapping between RDF triples, RDF quads, and CSV/TSV files. We developed an evaluation strategy for the format conversion and mapping functionality using so-called round trip tests.