MLQ framework

Main contributions (detailed results):

• • We empirically demonstrate the importance of the test set size to assess quantifiers

  • We show that current quantifiers generally have a mediocre performance on the smallest test sets

  • We propose a meta-learning scheme to select the best quantifier based on the test size that can outperform the best single quantification method

We also built a R Package, including all quantification methods used in our experiments as well as the auxiliary functions. Moreover, all our source code and datasets are available on the gitlab repository link