Character-Based Data-to-Text Generation

Codebase for the paper "Copy mechanism and tailored training for character-based data-to-text generation" (Roberti et al., ECML-PKDD 2019).

Step-by-step guide

Requirements

A working Python 3 environment is needed. Required libraries are listed in the requirements.txt file, use one of the following commands to install them, depending on your environment:

pip install requirements.txt
# XOR
conda install --file requirements.txt

Training

The main.py file is used to train an EDA_CS, EDA_C or EDA model on the dataset on your choice:

python3 main.py --dataset <dataset> --model <model>

The default configuration trains EDA_CS on the E2E+ dataset. Available models are ['e2e+', 'e2e', 'hotel', 'restaurant']; available datasets are ['eda_cs', 'eda_c', 'eda'].

Different hyperparameters can be set via argparse (run python3 main.py -h for more details).

At the end of the training phase, one checkpoint for each epoch will be stored in the trained_nets/<timestamp>/ folder, where timestamp is the UNIX time of starting the script.

Generation

The create_eval_files.py script will generate both outputs and references files, which can be directly used as inputs for the evaluation script. For example, you can generate on the E2E development set using ED+ACS as follows:

PYTHONPATH=. python3 utils/create_eval_files.py trained_nets/<timestamp>/<checkpoint> --subset dev

This will create the trained_nets/<timestamp>/<checkpoint>.dev.output and trained_nets/<timestamp>/<checkpoint>.dev.references files.

The default configuration uses your EDA_CS checkpoint to generate from the E2E+ test dataset's inputs. You can choose a different dataset/subset/architecture via argparse.

Evaluation

We took advantage of the E2E NLG Challenge Evaluation metrics. Please refer to their repository for detailed instructions.

Citations

Please use the following BibTeX snippet to cite our work:

@inproceedings{Roberti2019,
  author    = {Marco Roberti and
               Giovanni Bonetta and
               Rossella Cancelliere and
               Patrick Gallinari},
  title     = {Copy Mechanism and Tailored Training for Character-Based Data-to-Text
               Generation},
  booktitle = {Machine Learning and Knowledge Discovery in Databases - European Conference,
               {ECML} {PKDD} 2019, W{\"{u}}rzburg, Germany, September 16-20,
               2019, Proceedings, Part {II}},
  pages     = {648--664},
  year      = {2019},
  crossref  = {ECMLPKDD2019-2},
  url       = {https://doi.org/10.1007/978-3-030-46147-8\_39},
  doi       = {10.1007/978-3-030-46147-8\_39},
  timestamp = {Mon, 15 Jun 2020 17:05:23 +0200},
  biburl    = {https://dblp.org/rec/conf/pkdd/RobertiBCG19.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}
@proceedings{ECMLPKDD2019-2,
  editor    = {Ulf Brefeld and
               {\'{E}}lisa Fromont and
               Andreas Hotho and
               Arno J. Knobbe and
               Marloes H. Maathuis and
               C{\'{e}}line Robardet},
  title     = {Machine Learning and Knowledge Discovery in Databases - European Conference,
               {ECML} {PKDD} 2019, W{\"{u}}rzburg, Germany, September 16-20,
               2019, Proceedings, Part {II}},
  series    = {Lecture Notes in Computer Science},
  volume    = {11907},
  publisher = {Springer},
  year      = {2020},
  url       = {https://doi.org/10.1007/978-3-030-46147-8},
  doi       = {10.1007/978-3-030-46147-8},
  isbn      = {978-3-030-46146-1},
  timestamp = {Mon, 27 Dec 2021 15:13:42 +0100},
  biburl    = {https://dblp.org/rec/conf/pkdd/2019-2.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}