- Uncertain reasoning
High level reasoning and learning in the real world both involve processing information that is uncertain, imprecise and vague. Our research develops theories and methods based on mass assignments, random sets, fuzzy sets and probability distributions. These are applied to applications in information processing, machine learning, data mining, user modelling and logic programming.
Staff involved: T.P. Martin,
- Bio-mimetics and Sensors
Bio-mimetics is an engineering tool that covers a huge scale from the tangible nano-technology level, including materials, actuators and sensors, right up to the intangible level of artificial intelligence, including high-level perception, cognition and reasoning.
- Soft Methods in Knowledge Representation and Processing
The goal of artificial intelligence is to mimic human intelligence, and it is clear that people do not generally think using crisply defined categories. Instead, we use natural language where meaning is usually conveyed without requiring watertight logical definitions. Our work in soft methods for knowledge representation aims to model this uncertainty, with application in areas such as smart text processing, fusion of heterogeneous semi-structured sources, soft concept hierarchies and the semantic web.
- Linguistic modelling and computing with words
The paradigm of computing with words is of central importance in artificial intelligence and data mining since it facilitates inference from natural language rules and facts. Our research in this area has focused on the use of an alternative interpretation of linguistic constraints based on the random sets. This framework can be used as a basis for data mining algorithms which produce transparent linguistic models of the underlying systems.
- User modelling
The next generation of consumer goods will be much more sophisticated in order to cope with a less technologically literate user base. A user model is an essential component for "user friendliness", enabling the behaviour of a system to be tailored to the needs of a particular user. We investigate the induction of prototypical user models which can be represented in a number of different frameworks including conceptual graphs, Bayesian networks and object-oriented hierarchies.