BoxEL-based Query Relaxation in RDF Databases

The increasing adoption of Resource Description Framework (RDF) and knowledge graphs enables querying over large-scale, semantically rich datasets. However, users often struggle to formulate queries that return meaningful results because their queries are too restrictive. As a result, some queries return no answers, even though they are syntactically and semantically correct. Query relaxation techniques address this issue by rewriting queries that return no answers to retrieve additional relevant results. Ontology-based methods struggle to capture patterns that are not logically inferred from the knowledge base, even if such patterns occur frequently and could be learned.

This thesis proposes a novel query relaxation framework utilizing BoxEL embeddings, a geometric representation of ontologies based on axis-aligned hyperrectangles. By leveraging BoxEL's ability to capture hierarchies of concepts, the proposed method aims to enhance RDF query relaxation. The framework will be evaluated against existing query relaxation strategies using benchmark datasets. The expected outcome is an efficient, semantically meaningful approach to query relaxation in RDF knowledge graphs.