On Dependencies in Knowledge Graph Construction

Knowledge graph construction (KGC) requires numerous assets, such as shapes or mappings, to interact correctly.
However, maintenance and assessment of the quality of the pipeline implementing the construction is difficult, as
software engineering analyses and quality measures do not address the technologies used in KGC. In this paper,
we propose a syntactic, easy to compute notion of dependencies between assets, and show its capability to assess
change propagation. Furthermore, we discuss potential to use it for coupling and impact estimation. We evaluate
our approach using a prototypical implementation and a case study from the literature, where we find two bugs
where missing dependencies indicated an error due to miscommunication during change propagation between
the developers of two different assets.