This class implements the DFRee weighting model. DFRee stands for DFR free from parameters.
In particular, the DFRee model computes an average number of extra bits (as information
divergence) that are necessary to code one extra token of the query term with respect to
the probability distribution observed in the document. There are two possible populations
to sample the probability distribution: considering only the document and no other document
in the colection, or the document considered as sample drawn from the entire collection
statistics. DFRee takes an average of these two information measures, that is their inner product.