The sum in the equation is used to calculate the total distance between all combinations of publications in the sets A and B. It is expected that the distance is estimated using semantic similarity measures on the full-text of the publications, such as with cosine similarity on tf-idf document vectors.