Linguistic modelling based on semantic similarity relation among linguistic labels

In classical linguistic modelling for imprecise concepts and uncertain reasoning, a basic idea is to use fuzzy sets to describe the semantics of imprecise concepts, the theoretic foundation of this modelling is fuzzy logic. But in some practical applications, the determination of membership functions associated with vague concepts is difficult or impossible. In this paper, we present a new linguistic modelling technique based on the semantic similarity relation among linguistic labels. In this new linguistic model, a fuzzy relation is utilized to represent the semantic relation among linguistic labels. The elements in the fuzzy relation are interpreted as the degrees of similarities or the degrees of semantic overlapping between the corresponding linguistic labels. We develop an inference method for computing the degree of similarity between any two linguistic expressions which are the logic formulas generated by applying the logic connectives to the linguistic labels. This inference method utilizes the consonant mass assignment functions induced from the fuzzy similarity relation on linguistic labels. We show that the linguistic model presented in this paper satisfies some perfect properties, e.g. the excluded middle law, law of non-contradiction. This new linguistic modelling is illustrated finally in a linguistic decision-making example.