Efficient Inference and Computation of Optimal Alternatives for Preference Languages Based On Lexicographic Models