Optimal experimental design criterion for discriminating semiparametric models