Learning Lexicographic Preference Trees From Positive Examples