This is a Pytorch implementation of Reading Like {HER}: Human Reading Inspired Extractive Summarization".

python 2.7
pytorch 1.3
pyrouge 0.1.3

We evaluate our models on three datasets: the CNN, the DailyMail and the com- bined CNN/DailyMail (Hermann et al., 2015; Nallapati et al., 2016). You can download dataset from here.

We use the bandit settings and pretrained vocab embeddings provided by Dong et al. (2018). and you can download here.

Put all the downloaded files under folder data.

python dataLoader.py

python main.py / python evaluate.py --std_rouge