This repo contains code for Data-to-text Generation with Macro Planning (Ratish Puduppully and Mirella Lapata; In Transactions of the Association for Computational Linguistics (TACL)); this code is based on an earlier (version 0.9.2) fork of OpenNMT-py.

@article{puduppully-2021-macro,
  author    = {Ratish Puduppully and Mirella Lapata},
  title     = {Data-to-text Generation with Macro Planning},
  journal = {Transactions of the Association for Computational Linguistics},
  year      = {2021},
  url       = {https://direct.mit.edu/tacl/article/doi/10.1162/tacl_a_00381/101876/Data-to-text-Generation-with-Macro-Planning},
  pages = {510--527},
  volume = {9},
}

All dependencies can be installed via:

pip install -r requirements.txt

The main branch contains code to generate macro plans from input verbalization. The code for training summary generation is in summary_gen branch for MLB and summary_gen_roto for RotoWire dataset.

The test outputs and trained models can be downloaded from the google drive link https://drive.google.com/drive/folders/1jJjq5IvuBKNLTAe7fuwlDYParrxpK-WD?usp=sharing

The steps for training and inference for RotoWire dataset are given in README_RotoWire, and for MLB dataset are given in README_MLB.