Ranking Agent Statements for Building Evolving Ontologies
Ronald Siebes, Frank van Harmelen · 2002
In this paper a methodology is described for ranking information received by different agents, based on previous experience with them. These rankings again are used for asking the right questions to the right agents. In this way agents can build up a reputation. The methods in this paper are strongly influenced on human heuristics regarding the assignment of confidence ratings on humans. The methods provide a solution to current problems with ontologies: (1) handling contradicting and sloppy information, (2) efficient network use instead of broadcasting information and (3) dealing with ontological drift.