Get started
Work

Semantic data management with applications in data-driven chemistry

Year: 2026
Type: dissertation
Abstract: Scientific and industrial environments increasingly span experimental, analytical, and production-oriented processes, bringing together data that differs in format, structure, provenance, and level of abstraction. Conventional data architectures provide important foundations for scalable storage and processing, but they do not resolve the semantic heterogeneity that limits data access, integration, and interpretability. This dissertation argues that treating semantics as a core architectural concern is the key to unlocking the analytical potential of heterogeneous data. This argument is developed in four connected stages. First, a conceptual framework for characterizing data architectures is established and their suitability for heterogeneous industrial environments is evaluated. Second, the Semantic Data Reservoir (SEDAR) is introduced as a semantic data platform integrating semantic data access and automated analytics orchestration. Third, a formal framework for annotating, relating, and querying heterogeneous data sources in terms of domain knowledge is developed. Fourth, the potential of automated machine learning and large language model interaction to lower the barrier for non-technical users performing complex data-management and analytics tasks is demonstrated. The proposed concepts are realized in SEDAR and evaluated through application scenarios from chemistry. The results show that ontology-based semantics and workflow automation can improve the usability of data-driven research. They further show that semantic infrastructure, combined with automated machine learning, enables competitive performance on domain-specific tasks without extensive manual model engineering.
DOI: 10.18154/rwth-2026-08820
Language: English
Open Access status: green
OpenAlex ID: W7222255490
Cites:
Cited by:
Related to: