Multi-relational learning for recommendation of matches between semantic structures

Abstract

The paper presents the Tensor-based Reflective Relational Learning System (TRRLS) as a tensor-based approach to automatic recommendation of matches between nodes of semantic structures. The system may be seen as realizing a probabilistic inference with regard to the relation representing the 'semantic equivalence' of ontology classes. Despite the fact that TRRLS is based on the new idea of algebraic modeling of multi-relational data, it provides results that are comparable to those achieved by the leading solutions of the Ontology Alignment Evaluation Initiative (OAEI) contest realizing the task of matching concepts of Anatomy track ontologies on the basis of partially known expert matches. © 2013 Springer-Verlag.