The University of Sussex

MML, a modelling language with dynamic selection of methods

Vicente Guerrero-Rojo

Second generation knowledge based systems (KBS) often incorporate multiple problem solving methods. However the decision about which method to use is very much an open problem. The control involved in the dynamic selection of methods is considered a complex activity that requires the acquisition of specific knowledge and strategies. There is a need for modelling languages capable of handling, invoking, evaluating and choosing multiple methods at run-time. There are several modelling languages with such capabilities. With them it is possible to develop robust, more flexible and less brittle systems. Unfortunately, those languages are not flexible enough to cope with the behaviour of the systems when more methods are incorporated. In this paper we propose a new modelling language which overcomes these shortcomings. In doing this a framework is provided for reviewing the flexibility of current modelling languages.


Download compressed postscript file