Funding Provenance Sensemaking Visual Analytics

Behaviour analytics for defence and security (2019, £80,000)

Funder: Defence Science and Technology Laboratory (Dstl), UK

Date: 2019

Funding: £80,000

Partner: MASS Ltd

In this project, we use the human-machine teaming approach to design a new visual analytics system that supports kill chain elicitation:

  • The system allows user to mark up text in intelligence reports that may form part of a kill chain;
  • The system uses machine learning to learn from user annotation and recommend new information that may be part of a kill chain;
  • The analyst then comments on the recommendations indicating which is relevant and which is not;
  • The machine learning model uses the feedback to improve its model, which lead to better recommendation;
  • This process is iterative, so the system keeps improving as the user provides more feedback (this is also known as ‘active learning’);
  • The system can also provide explanation of why recommendation is made (‘explainable AI’), and the analyst can directly comment the explanation to instruct the model how to improve.