ORM ontologies with executable derivation rules to support semantic search in large-scale data applications

Abstract

A semantic layer maps complex enterprise data into an ontology with abstract business concepts that are well-known to business users. Chief data officers invest significant effort to create and update these ontologies, while data scientists do feature engineering by combining already existing concepts and features of the domain. However, it is a significant challenge to catalogue and maintain the numerous features pertaining to an ontology, which leads to duplicated features and unnecessary complexity. In this work, we propose to combine ontologies captured using the Object-Role Modeling notation with derivation rules defined in a datalog-like language called Rel, which allows the creation of a semantic layer with feature search capability. Our prototype framework uses the RAI Knowledge Graph Management System, which provides automated and incremental derivation rule evaluation.