How to capture the relationships between data? The DEK-G Methodology for designing an Enterprise Knowledge Graph

Abstract

Organizations struggle with the increased volume of data, as obtaining relevant information or discovering patterns becomes more difficult. One of the ways to address this problem is by creating a unified view of information, independent of the data sources. This can be achieved by modeling the domain knowledge from the enterprise in the form of a graph. This domain knowledge is represented in the domain knowledge schema, which is the core element of the Enterprise Knowledge Graph (EKG). Limited research has been conducted to create an EKG, and even less research focuses on exploiting the structure of the domain knowledge schema. Moreover, these methodologies do not explicitly focus on the role of the organization. Therefore, this thesis proposes the DEK-G Methodology to support organizations to create an Enterprise Knowledge Graph (EKG). The designed EKG can answer complex business questions from various data sources and discover patterns in the data by applying graph algorithms. Practical guidance was provided by applying it to two case studies. For each use case, the DEK-G Methodology could be applied to create an EKG, which could answer complex business questions. Moreover, using various graph algorithms applied to the designed EKG, additional insights could be derived related to the structure of the graph. It was shown that the organizational context influences the design of an EKG, in which the exploratory organizational context enabled more exploitation.