Search site

Artificial Intelligence

Our research within this theme exploits synergies between several key approaches to artificial intelligence. In computer vision we have pioneered a broad ‘cognitive’ approach and focus on video analysis and segmentation, with application in many sectors, including medicine, security, sport and entertainment. We have been in active in this area for over 20 years, with many alumni in academic and industrial positions around the world.

In knowledge representation and reasoning (KRR) we focus on representing information in a high-level way that allows integration of information from very different sources and the generation of new knowledge by reasoning from existing knowledge. The underlying data are produced in many ways: cameras generate video images, web browsers generate text documents (as well as images and sounds), wireless sensor devices generate numerical measurements of temperature etc. We make use of specialist techniques from areas such as computer vision and natural language processing have developed to extract information from these different kinds of data. This leads to three main areas of expertise: (i) qualitative spatial reasoning including both the underlying mathematical theories and the applications in areas such learning from video and geographical information, (ii) user modelling and personalization, and (iii) interpretation and fusion of spatial sensor data and integration with other sources of geo-spatial information. Running across all these areas is our work in the application of semantic technologies using ontologies as a tool to describe the content and organization of knowledge in different domains.

"In Natural language processing (NLP) our contributions include sentiment analysis and emotion recognition, Arabic language modelling, and Statistical algorithms for figurative language recognition. We share our expertise beyond the NLP community we share our expertise beyond the NLP community, and collaborate on interdisciplinary projects involving language engineering and with potential applications in textual data mining for detecting terrorist activities."

If you are interested in joining us for PhD study or in a post-doctoral position, you can find out about the opportunities that are currently available.

Examples of the importance and impact of our research:

Qualitative spatial reasoning Qualitative spatial reasoning is about handling information which can include concepts such as next-to, inside, and part-of rather than geometrically specific data with numerical coordinates. The Region-Connection Calculus (RCC), largely developed in work at Leeds over the last 20 years is today one of the most widely used calculi for qualitative spatial reasoning. Extensions of this research are currently being used to extract knowledge from videos with applications to surveillance.

MTU Mapping the Underworld Buried pipes and cables are an expensive problem when you need to dig up the road. How do you know what's there before you start digging? Can you rely on finding paper records that can be decades old, and can you detect what's there without digging? We have played a major part in a series of funded projects that have (1) addressed the problem of integrating utility records – with a live implementation now in Scotland (which won an IET Innovation Award in 2012 and a 2012 NJUG Avoiding Damage Award) and we are now (2) fusing outputs from a multi sensor device to map buried utility apparatus.

Ordnance Survey MapWorking in collaboration with The Ordnance Survey's Geosemantics team we have developed an intuitive tool, called ROO to support domain experts (ecologists, geographers, emergency planners, etc.) without knowledge engineering skills to create and maintain a conceptual ontology.