Collaborative filtering guided exploration in multi agent filtering framework

In order to improve a quality of filtering services, an exploration that is based on a usage of knowledge, being hidden in numerous already processed requests, has to be performed. Small scale applicability and ignoring a necessity to control the exploration are major drawbacks for many existed systems which try to explore new information areas. The essence of the proposed collaborative guided exploration is in its comprehensive control over a level of a request adaptation and in its applicability on huge spaces, where filtering requests are defined. The applicability of the presented exploration is illustrated in one travel finder system (URLAUB) and in one personal information assistant (PIA). Experimental results show that filtering requests, for which one good neighbourhood exists, can significantly be improved by adding new relevant words.