Towards Data Analytics Using Contextualized Knowledge Graphs

A contextualized knowledge graph (CKG) allows for the representation of refined information organized into specific contexts and along multiple contextual dimensions, e.g., time, location, and topics. Although existing knowledge graph management systems offer robust capabilities for data modeling and reasoning, these platforms often lack native support for performing analytics under the considerations of the characteristics of big data, particularly in domains that require real-time predictive modeling and graph-based machine learning. This paper presents a concept for an extension to a cloud-native KG Lakehouse architecture that integrates scalable feature engineering and machine learning capabilities over CKGs. We introduce a distributed analytics layer to the KG Lakehouse that supports predictive data analytics over CKGs. Context-aware operations for Knowledge Graph OLAP will facilitate the extraction of features while remaining agnostic to specific contextualization strategies. The result is a general-purpose architecture that enables scalable and semantically grounded analytics on dynamic contextualized knowledge graphs. The proposed extension is modular and cloud-native by design. Implementation and evaluation remain future work.

Keywords: Contextualized Knowledge Graphs, Big data analytics, Data lakehouse, Machine Learning, Cloud-Native Systems