This repository contains the dataset and code for the INLG 2019 paper Revisiting Challenges in Data-to-Text Generation with Fact Grounding

🤝Please kindly cite this work if it helps your research:

@inproceedings{wang-2019-revisiting,
title = "Revisiting Challenges in Data-to-Text Generation with Fact Grounding",
author = "Wang, Hongmin",
booktitle = "Proceedings of the 12th International Conference on Natural Language Generation",
month = oct # "{--}" # nov,
year = "2019",
address = "Tokyo, Japan",
publisher = "Association for Computational Linguistics",
url = "https://www.aclweb.org/anthology/W19-8639",
doi = "10.18653/v1/W19-8639",
pages = "311--322"}

Get the dataset

• 👌The dataset/scripts contains the scripts to prepare the dataset from stratch.

  • The crawl, purification and enrichment directories have respective README.md files inside to describe their functionalities.

• 👍Download from rotowire_fg and place the 3 folders under dataset/

  • The final purified, enriched and enlarged dataset are included in new_ncpcc without having to execute the scripts.

Run the model

👉Please go to the README in under model

Run Evaluation Script

👉Refer to this line in the model/run.sh