An Agile and Ontology-based Meta-Modelling Approach for the Design and Maintenance of Enterprise Knowledge Graph Schemas

Abstract

Enterprise Knowledge Graphs (EKGs) are increasingly created and used by organizations for structuring knowledge of a particular application domain and consequent exploitation through analysis, reasoning and integration of information extracted from different data sources. Yet, one of the main challenges is designing and maintaining EKG’s schema, which requires high expertise in ontology engineering and the addressed application domain. Various approaches and tools offer visual aids, but they target ontology engineers and neglect the domain experts. Domain-specific modelling languages (DSML), in contrast, offer concepts that domain experts easily understand because of their tailored graphical notations. DSMLs can be created with a meta-modelling approach, which does not require ontology expertise but can be adopted for the equivalent creation of EKG schema. This paper presents an approach that extends the traditional and sequential meta-modelling approach with an agile one, allowing domain-specific adaptations of modelling languages and testing them on the fly. In this way, the domain experts are facilitated to be in the engineering loop. Moreover, the approach foresees automatic mechanisms to ensure that an EKG schema is designed while performing the visual domain-specific adaptations. The approach has been developed by following the Design Science Research methodology, which led to the creation of a prototypical tool called AOAME. The latter has been used to implement real-world scenarios to evaluate the proposed approach’s utility. The correct design of the approach has been evaluated by tracing the prototype functionalities back to the requirements.