Qualitative modeling with gradual and similarity rules

A model for qualitative reasoning with linguistic values is proposed. Linguistic variables in propositions contain qualitative (linguistic) descriptions of system attributes. The meaning of the descriptions is given as a possibility distribution defined on the corresponding universe of discourse. Interactions among system variables are specified by means of gradual and similarity rules. We give the structure of the rules, pattern matching and conclusion computation. Evidences resulting from rules are combined by means of several evidence combination methods, and propagated further to the next set of possibly applicable rules. Since there is no restriction on the domain and range of rule mappings, the dynamics of the system is easily modeled. We demonstrate this on the example of variables used in scheduling, by representing the domains of system variables as nodes of the graph, and rules as links among them. We call this structure linguistic causal graph