quantifying relevance is difficult at best, so building test data is at least as difficult. If there is art in defining the score in the test data, a fuzzy alg

quantifying relevance is difficult at best, so building test data is at least as difficult. If there is art in defining the score in the test data, a fuzzy algorithm is a necessity.