Comparison of fuzzy subsets: towards a linguistic approach