We release the data and codes used in our paper.
We uploaded the core codes in File "data":
"yelp" contains synthetic semi-structured training pairs and generated summaries of Yelp.
"amazon" contains synthetic semi-structured training pairs and generated summaries of Amazon.
"rt" contains synthetic semi-structured training pairs and generated summaries of RottenTomatoes.
We uploaded the core codes of synthetic data creation and proposed models in File "code". The details are described in code/README.md.
Both basic model and advanced model are implemented based on Fairseq framework provided.