This repository shares the code and data for our paper Script parsing with hierarchical sequence modeling for interest of reproducing its results. For questions, please mail to Zhai Fangzhou.
The core enrironment is AllenNLP 1.0 with Python 3.7. For further details, please refer to ./wp120200722
All data is included in ./data_combinations, including that from InScript, DeScript and back-translation paraphrased InScript, and our train/val splits. Our experiments were conducted with different combinations of these files.
- your data format should follow that of the files in ./data_combinations
- adjust settings in SSP_optimization_script.py
- execute SSP_optimization_script.py