Data Intelligence builds the bridge between methods for handling data and application domains. The methods for handling data include those for converting data to knowledge (e.g. machine learning, deep learning, word embedding) and those for interacting with data and knowledge (e.g. semantic search, chatbots, recommender systems). These methods are generic and can be applied to any data. Adapting these methods to the specific requirements and way of working in various application domains is not straightforward - it is easy to create a solution which gets in the way, but difficult to create a solution which slips seamlessly into established workflows to subtly support knowledge workers. Challenges include efficiently creating training data for a learning system (avoiding the need for expensive experts to spend too much time labelling data); and evaluating how well a system is performing in terms of efficiency and effectiveness, in order to get an objective measure of the increase (or decrease) in productivity though using the system.
Many of these challenges throw up important scientific questions to be solved in the Data Intelligence Lab. We will also concentrate on transferring scientifically-validated best-practice solutions to industry.