GovWire

BMT Defence Services, working with Oxford University and Rescue Global, received funding from the Centre for Defence Enterprise (CDE) to develop a way to extract actionable intelligence from a range of unstructured data sources to enhance decision making in the field.

Making the best decision in a short time is dependent on being presented with a single view of the unfolding situation with a confidence level which is derived from the wealth of uncertain data available.

This could help with a disaster response scenario, for example, where many data sources exist (satellite imagery, open source reports and professional monitoring services).

The project has shown that a crowd of volunteers can identify features (flood, damage and tarpaulin) in unstructured datasets (in this case satellite imagery) with similar performance to a group of experts. It has also shown that the machine learning algorithms can estimate the accuracy and bias of the crowd members; and by assessing that accuracy, it enables the intelligent assignment of tasks to users to reduce the time taken to produce a consolidated view.

For the next phase of work, the team will scale up and extend the current approach to produce a platform that addresses 3 key challenges; the introduction of intelligent agents alongside the crowd, automated and targeted training of both the crowd and finally management of the evolution of the information generated with a scalable and queryable provenance data store.

Simon Luck, Department Head of Information Systems, BMT Defence Services says:

CDE funding has provided us with the opportunity to deliver Research & Development activities alongside academia and non-for-profit organisations to prototype revolutionary capability.

About CDE

CDE funds novel, high-risk, high-potential-benefit research. We work with the broadest possible range of science and technology providers, including academia and small companies, to develop cost-effective capabilities for UK armed forces and national security.

CDE is part of Dstl.

View the original news story