Problem in b_cubed calculation about metric4coref HOT 1 OPEN

Thanks, this line is an unnecessary intersection indeed.
But exactly, it is the intersection of two same mention set.

For example, if predicted_clusters is [["a", "b", "c"], ["d", "e", "f", "g"], ["h", "i", "j"], ["k"]]

At the calculate of sum(predicted_clusters, []), we will get the original mention list ['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k']. It is the same process in gold_clusters.

All the mentions should exist in one of the predicted_clusters or gold_clusters. So the intersection is unnecessary.

I think in this line, we should add assert. It will be more clear.

assert set(sum(predicted_clusters, [])) == set(sum(gold_clusters, []))

What do you think about this?

from metric4coref.