Dataset
Explore
Download
Construct the Dataset
We provide
scripts
to recreate the dataset. Please follow the provided instructions in the README.
Evaluation
We provide a
ROUGE
evaluation
script
. We use F1 ROUGE to evaluate the quality of the full generated answer and the final 20% of the answer.
Pretrained Models
Seq2Seq Multi-task
Abstractive
Pretrained sequence-to-sequence model to abstractively generate answers.