FaithEL: Strongly TBox Faithful Knowledge Base Embeddings for \(\mathcal{E}\mathcal{L}\)

Knowledge base embeddings map both TBox and ABox axioms into vector spaces. Recently, there have been various approaches for embedding normalized \(\mathcal{E}\mathcal{L}\) knowledge bases (KBs). However, none of these approaches show whether TBox axioms of a given KB match those that hold in the embedding. Embedding models that exactly match with the TBox part of the KB are called in the literature strongly faithful TBox models. Here we present an implementation of a strongly faithful TBox embedding method for normalized \(\mathcal{E}\mathcal{L}\) knowledge bases.