Hi colleagues,
very painful this usage of the unit test avoiding any debugging.
I am stuck with the rouge1_similarity. It passed the first test but failed for the 3 others.
my counters are correct I guess, my formulas for precision and recall are consistent as I am using the sum of the tokens counts for the denominator (sys for the precision and ref for the recall) and my f1-score formula is also valid.
So I need your feedback about this method.
Thanks
Jerome
