Adaptive Agent-based Systems for the Web:An Application to the NECTAR Project

The paper presents an overall framework for devising adaptive Web systems characterised by a smart interface that exploits personalised anthropoid characters. Part of the framework is currently investigated for the NECTAR system, a smart interface for on-line shopping. We propose an agent architecture and highlight the roles of different agents that can carry out tasks required for a smart interface. We define knowledge structures and reasoning abilities that can be considered by such adaptive systems. In particular, we point out how adaptive Web applications could benefit from using case-based reasoning, data mining, and knowledge discovery techniques.