The project is structured into four work packages (WPs), as illustrated in Figure 1, which together address the project objectives ranging from understanding the impact of Knowledge Graph (KG) characteristics on KG Embedding (KGE) methods to empirical evaluation, engineering of characteristic-aware KGE methods and benchmarking frameworks. The work packages are designed to progress from conceptual analysis to empirical validation and practical engineering, ensuring both scientific rigor and real-world impact.
WP1 establishes the conceptual foundation through a systematic literature review of Knowledge Graph characteristics and their influence on Knowledge Graph Embeddings. Building on this, WP2 develops an empirical evaluation framework, including controlled experiments, benchmarks, and KG profiling to quantify the impact of KG characteristics. WP3 investigates and advances KGE methods by analysing their behaviour under different graph conditions and proposing characteristic-aware embedding approaches. WP4 focuses on knowledge engineering and algorithm design, translating research findings into practical guidelines for knowledge engineers.