Ontology Learning from Relational Database: How to Label the Relationships Between Concepts?

Developing ontology for modeling the universe of a Relational Database (RDB) is a key success for many RDB related domains, including semantic-query of RDB, Linked Data and semantic interoperability of information systems. However, the manual development of ontology is a tedious task, error-prone and requires much time. The research field of ontology learning aims to provide (semi-) automatic approaches for building ontology. However, one big challenge in the automatic transformation, is how to label the relationships between concepts. This challenge depends heavily on the correct extraction of the relationship types. In fact, the RDB model does not store the meaning of relationships between entities, it only indicates the existence of a link between them. This paper suggests a solution consisting of a meta-model for the semantic enrichment of the RDB model and of a classification of relationships. A case study shows the effectiveness of our approach.