This repo contains code for Data-to-text Generation with Macro Planning (Ratish Puduppully and Mirella Lapata; To appear: In Transactions of the Association for Computational Linguistics (TACL)); this code is based on an earlier (version 0.9.2) fork of OpenNMT-py.
All dependencies can be installed via:
pip install -r requirements.txtThe main branch contains code to generate macro plans from input verbalization. The code for training summary generation is in summary_gen branch.
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.