Ideally, the comparison between two sets of results should be handled by a Comparer function, at its simplest calculating the difference between two scores for two combinations of texts/variables and creating a sort of difference-score.
It would also be nice to have a rank-comparer: e.g. if a token is ranked 1st w.r.t. relevance to a certain label, and 10th w.r.t. relevance to another label, the rank-comparer should output something like "-9".
Ideally, the comparison between two sets of results should be handled by a Comparer function, at its simplest calculating the difference between two scores for two combinations of texts/variables and creating a sort of difference-score.
It would also be nice to have a rank-comparer: e.g. if a token is ranked 1st w.r.t. relevance to a certain label, and 10th w.r.t. relevance to another label, the rank-comparer should output something like "-9".