A basic aim of the development of fuzzy linguistic models is to produce fuzzy systems which have both a high accuracy rate and a high degree of transparency. This paper presents a modelling method which allows the creation of accurate fuzzy linguistic models, based on fuzzy subsethood-values. A resulting model is represented in the form of weighted fuzzy general rules, employing relative weights generated from fuzzy subsethood values. These weights are adjustable according to the datasets available for learning. The effectiveness of this work is demonstrated with experimental comparative studies.