Time-Based Modeling of Linguistic Preference to Preferential Probability

Abstract

In this paper, we present a method to estimate the likely concept a committee of designers will select given their verbalized preferences toward each alternative. In order to perform this estimation, we present a new method of preference elicitation based on natural language. First, we show a way to model preference in the natural language of appraisal, which describes the degree of intensity and the uncertainty of preference based upon gradable semantic resources to express appraisals. We then show a way to map linguistic appraisals into probability distribution functions. Finally, we present a Markov model that utilizes these probability distribution functions in state transition matrices to calculate in a time-varying manner the change of preference over time. We present a case study to illustrate the validity of the method.