Download test datasets: https://github.com/harvardnlp/sent-summary
git submodule init
Run through the InferSent setup, downloading the Glove embeddings and the pretrained encoder.
Run the bootstrapped Gain analysis on the test data:
python infersent.py ./sumdata/Giga/input.txt ./sumdata/Giga/task1_ref0.txt