Approximation of sets of superwords by L-language formulas