Representing entities from an ontology as geometric shapes (such as balls, boxes, etc.) in a low-dimensional vector space, known as Region-based Geometric Knowledge Graph Embedding or RKGE, has demonstrated the ability to outperform traditional knowledge graph embedding methods in reasoning tasks while preserving the structural properties and syntactic characteristics of ontological axioms. In this study, we introduce a novel approach to enhance the subsumption capability of geometric embeddings based on n-balls. Additionally, we propose techniques to enhance the quality of such embeddings by extracting meta-information from the information-rich lexicons or annotations within the domain ontology.