A derived information framework for a dynamic knowledge graph and its application to smart cities

In this work, we develop a derived information framework to semantically annotate how a piece of information can be obtained from others in a dynamic knowledge graph. We encode this using the notion of a derivationand capture its metadata with a lightweight ontology. We provide an agent template designed to monitor derivations and to standardise agents performing this and related operations. We implement both synchronous and asynchronous communication modes for agents interacting with the knowledge graph. When occurring in conjunction, directed acyclic graphs of derivations can arise, with changing data propagating through the knowledge graph by means of agents' actions. While the framework itself is domain-agnostic, we apply it in the context of smart cities as part of the World Avatar project and demonstrate that it is capable of handling sequential events across different timescales. Starting from source information, the framework automatically populates derived data and ensures they remain up to date upon access for a potential flood impact assessment use case.

Automated summary

This work develops a framework to annotate how information can be derived from others in a dynamic knowledge graph. It encodes this using the notion of a derivation and captures its metadata with a lightweight ontology. The framework provides an agent template for monitoring and standardizing the process, and implements synchronous and asynchronous communication modes for agents interacting with the knowledge graph. It is applied in the context of smart cities and demonstrates the ability to handle sequential events across different timescales.