BoxLitE: A Faithful Knowledge Base Embedding Based on Convex Optimization
Researchers have developed BoxLitE, a novel knowledge base embedding model designed for DL-LiteH. This model leverages convex optimization to represent concepts as convex regions in a vector space, enabling better representation of hierarchical knowledge found in ontologies. BoxLitE aims to faithfully embed knowledge from both factual (ABox) and conceptual (TBox) components of knowledge bases. AI
IMPACT Introduces a new method for knowledge base embedding that could improve how AI systems understand and utilize structured knowledge.